Transcript: EP 01 — Airports: Can math price a flight delay?
6,847 words · 41:26 runtime · every timestamp jumps into the video.
0:09Thomas The podcast where we pit math versus vibes to answer the question that we happen to be thinking about at a certain time that week. Before we get into that, I think maybe we can let people know the shtick we have here, which is this is Math versus Vibes. Eric is Team Math. I'm Team Vibes. The game plan for this podcast is I'm gonna ask Eric a question that's been nagging me and I believe nagging most people in our world about things that occur in our society that are mostly vibe based. Like I think this might be happening because of that, because of this, who knows. Eric will then take that question and apply a very complicated, algorithmic, I don't even know, model using his computer science technology skill to analyze and find an answer to the question I asked. He will then present it to me and I'll ask him some questions and then we'll see if he even had my question right, if not the answer. We'll go back and forth and then ultimately we will understand and decide, is my vibe read on that question right or is Eric's math model reading the question right? And so this is our first time doing it and it is the first question we asked, which is this question about airports. And I do remember it, Eric, I believe, which was something. Yeah, it's something along the lines of a modern airport has so many variables in play. I don't even know how many gates and how many flights coming in and out and then you add in weather, refueling, the requirement for unexpected maintenance. How can you ever plan and have flights actually execute on time? I actually think airlines are doing a great job. I think airports and airlines are doing a great job because a delay is so rare it blows my mind. So I do have so many questions about airports, similar to grocery stores. I don't need to get into that, but food expires, like how do you know who's gonna get what? Incomprehensible variability, how do you create a system that can actually operate effectively and so you're not losing money all the time to make it happen? And yeah, the main, in the genesis of this was I had a flight delay and it went and then it flew right away as planned an hour later at a different gate. Another airline was in there. Infinite variability, humans, weather, famously unpredictable. How do we make it happen, Eric?
2:57Eric Let me hazard a guess. I think they're running it based on vibes.
3:02Thomas Yeah, so and we'll get into it. I think there is a certain peanut butter spread of just human like ah we're just gonna do it, this is gonna work. Like I don't think there's a system. I don't think there's a system, but I don't think the system is systematic as people think the system is. So I think it's, I think there is a system but I think that the vibes are in play and I think it's also not as predictable as we would like to think. So I think there's a system but it's not as influential as we would like to think, that's what I'm saying. So hence team vibes wins on this one as well. My stance currently going in.
3:44Eric Yeah, no that's good and I think I'll give you that. I mean I think there is often you know for all these complicated systems we have to run things. There is a lot of like tribal knowledge or tribal know-how that kind of ends up doing some of these final details that sometimes are hard to manage. There's a lot of that going on in the road.
4:06Thomas Seating
4:07Eric already,
4:08Thomas but yeah go ahead.
4:09Eric Yeah I just remember you mentioning a point on you know like when do these actual numbers actually matter and you know one of the things where I think they actually matter is when you're trying to schedule these flights. They're actually quite precise on when they plan to land, when they plan to take off, if they decide to have slack in between. So they might say it's scheduled to land at 10.10, it's expected to land at 10.05, gives you a little slack in case you run into some issues along the way.
4:38Thomas So just to be clear time, times or numbers all of a sudden?
4:44Eric Yeah I mean there certainly are one type of number.
4:47Thomas October 29, 2024.
4:50Eric And do you know why it's kind of a calm day? It's shoulder season. This is kind of fall, you know we're in a little bit of a lull. This is common. So we can get more into the statistics in a little bit here.
5:02Thomas Sorry I'm just asking questions.
5:05Eric No this is a good point because we only get six days. I gave us six samples. A little sneak peek. I'm just gonna give it a run through so you get a flavor of it and then we'll delve into some details all right. So here we're starting in the morning.
5:21Thomas These are all the flights that occurred on October 29th.
5:25Eric Yeah so
5:26Thomas it's so much more dense on the East Coast but yeah okay.
5:29Eric Well and it starts over there earlier you know. Yeah you kind of saw the kind of the flux at the beginning from the East Coast. I think we'll see something similar at the end where the West Coast kind of has
5:40Thomas some flights going on. What I'm looking at visually here is all of the largest airports and the circle size is based on volume of takeoff and landings and the blues are departures and the reds are landings. Is there a cold coordination for delays?
6:00Eric Yeah so and actually that's just on the bottom. So I kind of overused red and blue just because that's you know that's our color theme. So you know I'm trying to build up some nice vibes here. So actually it's kind of confusing but on the bottom sparkline blue is departure, red is arrival. In the actual graphic most of this is styled based on delays. So if you look up here and I zoom in this airport's experiencing some delays right now. So
6:30Thomas how can you tell that because of because
6:33Eric it's it's red
6:34Thomas okay.
6:36Eric And any flight that's red any flight that's red so like this guy is actually experiencing a delay it took off late.
6:44Thomas And so the entire flight is then considered delayed obviously right because they're not making up time for the most part.
6:51Eric Well that's the thing that's why you know I brought this up earlier you know we have scheduled the part of your scheduled arrivals it's it's common to actually include some slack time such that you might be delayed at takeoff five minutes ten minutes and make it there on time.
7:08Thomas Which
7:08Eric wooo wooo wooo
7:09Thomas vibe check. I feel like in the past six years they've increased the stated duration of a flight to therefore increase the slack because now I have like a flight to San Francisco it's four and a half hours and then you get on the plane like and we'll be landing in three hours is 59 minutes. And I'm good with that because it makes that like everyone's in but I feel like in the past and this is classic vibe language I feel like in the past six to seven years they've increased that slack.
7:40Eric I completely agree I feel the same way but when I look at the statistics it seems like it's not true somehow.
7:49Thomas Really so the the the stated flight duration for common routes has not expanded?
7:57Eric That's a good question I can't tell you the answer on that but I can show you statistics that lead me to believe that they don't have all this extra slack in the system. And so let's just talk about the economics right now all the extra slack you add is gonna decrease your revenues right?
8:15Thomas Well wait decrease your revenues right?
8:18Eric Decrease sorry yes. Oh
8:20Thomas yeah so it's less new seats that you can sell.
8:26Eric Yeah so you're basically reducing your utilization of your fleet right? Ideally you can utilize it 100% you take off and land exactly you have no extra slack maximize revenue right? But this is extremely brittle system right? If it if one mistake happens it's gonna cascade throughout the day like you said right?
8:47Thomas Well interestingly brittle and severe because the risk of failure can be catastrophic.
8:56Eric This
8:56Thomas will be another really interesting piece that we could bring up and obviously we can just go in rabbit holes all day long but that's what Southwest business model was was super fast turnaround time to maximize utilization
9:07Eric that's
9:08Thomas why there's like cleaners come on as people are de-boarding and stuff like that. And then it gets in the question of economics like what's more profitable multiple small short-range flight you do ten flights between Minneapolis and Memphis or four to New York and LA anyway that's not the vibe that's not the question we're asking we're talking about managing can you manage a complex a system as complex as this and I'm I'm seeing you have the data Eric but I'd like to answer the question.
9:43Eric Yeah well let's dig in a little deeper right I still don't think we quite understand how there's so much variability in the system so you know like very small things seem like it can just throw it in or to a ruckus but before we go into some of these nitty-gritty details let's step back and look at some of the trends because you know I alluded to this like I actually don't think it's true even though I kind of feel the same thing but I actually you know here let me see the trends why I said that was based on the delay over time so here's so like I said I gathered data all the way back from 2003 all the way to present so the the Transportation Department or BTS maybe
10:29Thomas Department of
10:29Eric Transportation yeah Bureau of Transportation I can you know I can't quite remember but they publish this great data set and it's extremely long so back to 2003 I went through and analyzed the statistics and looked at the delay so they have a scheduled departure scheduled arrival actual departure no way every
10:51Thomas commercial flight in the u.s.
10:52Eric yeah
10:53Thomas wow okay yeah
10:54Eric that's what we were seeing there for a single day so and I have that for literally every single day obviously we don't need to see it for every single day it's fine
11:04Thomas that's for the patreon that's for the that is for the three-hour long long-form pod yeah
11:08Eric if you're really digging this you know we'll go through it but I
11:12Thomas mean December 14th 2011 we're all we all want to know
11:17Eric yeah what
11:18Thomas was happening in the skies
11:20Eric yeah yeah well let's dig in but not right now okay let's stay focused here so anyways where I was going was the average delay back then was nine and a half minutes in 2003 so this is right when we were in high school
11:34Thomas yep way way
11:36Eric and
11:37Thomas okay walla walla yep
11:38Eric now the average delay is seventeen point three minutes and
11:44Thomas this is delayed departure
11:45Eric yeah okay so almost
11:50Thomas double
11:51Eric yeah when I think about this you know so I think a lot of the delays and we'll look at the statistics here in a second a lot of the days ends up being in the afternoon Joe just by and actually we'll look at it first here like so this is
12:05Thomas yeah
12:05Eric this is average delay by time of day so over the years so in the morning when it costs us 5 a.m. to noon the average is I don't know what is this five to maybe 11 in the midday you know we're up another two three four minutes on average and averages are really hiding a lot of it because most flights you know like you're saying just work and believe in when there's flight to the way yeah yeah our two hours wouldn't because
12:35Thomas you're right because these have such swings like you or like a so
12:38Eric our cancellation is the median the median delay is often probably zero I get 50% of the flights don't have delay it's gonna be zero
12:46Thomas yeah
12:47Eric what we're looking at is percentiles like 70th percentile 90th percentile and that's the snaps that I show in a lot of these here we're just showing mean to kind of show some general trends and that you know it's kind of shows this like muted response because you're averaging with all these things that just work yeah but you have these events that don't work right
13:08Thomas let me in and also Eric we're gonna go over a time but that's fine cuz I this is incredibly interesting what about cancellations are they not contemplating this at all
13:17Eric there is I don't have stats every time which is a miss
13:22Thomas what yeah noted ooh noted
13:25Eric so
13:26Thomas easy to be the vibe yeah where you can just attack all of your work but I think okay I'm with you see
13:32Eric if we if we go back here so in general when we show a single day of flights we show a completed percentile so like this day the full day ninety nine point eight percent of flights win that's
13:43Thomas incredibly just flat-out that's that's pretty incredible yeah
13:47Eric so 17,000 flights 31 are canceled crazy great 93% are on time so if you're looking at the median flight everything's zero right there's a
13:58Thomas non-weather day right yeah
14:01Eric this is clear sky day so we're just showing the system working and even when it's working you know you see these days because you know there's a few bad flights and they have a long delay right
14:12Thomas yeah
14:12Eric do
14:14Thomas we know the reason for those delays in good weather we can presume it's probably maintenance or staff actually is o3 so maybe they weren't all shortcut on staff but do they have a great
14:23Eric question which brings us to another part of the data set so they actually publish the cause of these delays you know it's
14:31crazy a
14:31Eric little crude but yeah I would have never expected all of this data
14:34Thomas yeah
14:35Eric so here's the categories they're very small on the bottom I'll read them it read is late aircraft uh-huh blue is carrier delay yellow is NAS ATC so some sort of code from air traffic controllers so they have some reason to delay a flight weather and security wait
14:56Thomas weather's um the green
15:00Eric it's yeah this small green one actually oh
15:02Thomas it's a security is not even visible cuz it's so rare
15:05Eric yeah I mean I maybe can kind of see a little green line but not I'm
15:10Thomas shocked that weather is such a small shot you know my I mean obviously that's at the core of all this that my my hiccup on weather that it's represents what
15:19Eric is kind of shocking but then so was there actually a weather issue at your airport
15:26Thomas well it suppose it delayed other there were other flights delayed that people told me was because of the weather
15:32Eric okay
15:33Thomas but but there was never an actual storm but there was a tornado watch there was like weather coming in this
15:40Eric thing eludes to it at the bottom you know like the big chunk is these late aircraft and these might be late because of weather yeah and what we're seeing is like the cascade ends up showing up in this late aircraft yeah your
15:54Thomas sections final bucket that yeah is the experience it's
15:59Eric like the carrier says it's late who knows why really right
16:03Thomas you know what's interesting I wonder who pays you know like what where's the reimbursement at because whether you know like the layer lines don't need to repay for a ticket late aircraft maybe they do
16:15Eric yeah there is laws on that stuff now huh
16:18Thomas yeah yeah and so I'm sure that the coding of this too is like very intentional up because it would represent like what the what the delay is caused by and what they have to pay for but okay my takeaway here is my comment about the weather being such a crazy factor I mean that's like what 5% of the of the cause though to your point late aircraft let's let's be generous and say late aircraft like 25% later craft is because of weather that's still not the major cause of delays weather is not the major cause of delays right I
16:57Eric honestly can't tell you the answer to that because
16:59Thomas of the pink this is the vibe this is my vibe so this is let me teach you how to let me
17:06Eric tell you I'm hesitant to say that's true just because it depends on how your attribute delays right and the data is potentially not granular enough to understand the root cause of all these delays did
17:23Thomas we find our core rub there's never gonna be enough granular data dude we got to make judgments we got to make calls which I think might be where this may could be leading we'll never know so I'm with you that like
17:37Eric I'm not ready to make a call yet okay I'm just gonna jump to the vibes let's not try to think about this deeply at all like why would we want to look at more numbers like I think I know the answer like why would it
17:54Thomas wait
17:55Eric dude let me just vibe dude
17:59Thomas yeah please all
18:01Eric right slow your roll dude we that's your problem and that's you're such a vibe person now I just wanted to take a quick snap back and take a look at your text message so you're sending this text message
18:15Thomas yeah
18:16Eric is they in the evening in the summer and a surprise surprise and that the other night surprise you're delayed like I would look at the statistics and say why I would be shocked if you're sending this on time it's like I'm
18:35Thomas in the zone but wait wasn't 12 to 5 the afternoon and out the evening I mean if we're gonna be using modeling so we have our our axes that's
18:44Eric a good point so this is mountain time the whole models in Eastern time so this is actually 536 Easter is
18:54Thomas this like me sneak scheduling a Saturday night thing now you're trying to do a little I'm
18:59Eric just saying the statistics are not on your side you're saying all yours is incomprehensible yet all the statistics say exactly what you have happened to you should happen to
19:12Thomas I chose to fly at the comprehensively known most delayed time
19:20Eric yes
19:21Thomas in the scene you
19:23Eric were complaining about it to me because you're like why did this happen to me and
19:28Thomas I don't think I I guess I didn't fully say this um my flight wasn't delayed
19:33Eric that's
19:37Thomas what I was other people's flights were delayed and I was like oh maybe mine will be but it wasn't it took off on time and it was it was um it was great and this is just for me right now can you go back to the trends
19:50Eric yes I mean you're just gonna drop that and just act like that's okay what do you mean you weren't delayed what do you mean what do you mean I'm not delay
20:02this I
20:03Thomas was on a lot of flights did
20:05Eric what do you know the particular flight you were at the airport oh
20:09Thomas no I had to buy a different flight okay so I was apt I wasn't only delayed I had to in the uber switch to a different flight because the one was delayed and then more uh-huh more weather was coming in later that night in the evening at 7 p.m. in the summer where I had to buy a different flight that got out of 3 p.m. I got out of the window I got out in the prime 3 p.m. summer window so I did have a flight delayed on one airline which I then canceled for credit and bought a new flight at 8 p.m. and then had to call and have them move me to a 3 p.m. plane
20:48Eric all right I think we're gonna have to pull this up here let's see let me go wait
20:53Thomas hang on hang on can I see the trends real quick because I had
20:55Eric yeah two
20:56Thomas questions on that so yeah what's the um the horizontal axis so
21:01Eric this is yours okay
21:06Thomas it's
21:06Eric by season yeah the red is the 90th percentile flight so the 90th percentile flight so like say every day you know it's like 10 out of the hundred flights has this delay so that goes from 25 minutes up to like 48 minutes and
21:26Thomas that's the worst it's the most like
21:28Eric the 10% worst flights got
21:30Thomas it got it got it so like the upper end okay and then there was the other what was the other trends there's one you flashed
21:37Eric season
21:39Thomas yes I don't know season season season so got it okay so summer red interesting the winter like wait this is crazy winter has the same delays from oh six to like oh nine
21:57Eric so like like the blue right so
22:02Thomas like
22:02Eric yeah they're actually the most flat of all of these it does seem and even in kovat but it might have just been the timing of kovat right yeah but yeah winter
22:15Thomas I guess winter storms you know there are storms and maybe more delaying storms because of the nature of snow and ice and temperature
22:27Eric so this might be one reason to believe that the
22:31Thomas my June
22:32Eric storm delays have not changed much so the airport practices you know are not responsible for as much of the winter delays as the summer ones which are kind of changing more over time that assuming the business practices are sometime somehow driving this which is kind of where my head was at well
22:52Thomas also technology or can't we make some better snowplows
22:58Eric yeah so maybe that's counteracting some of these business practice changes the snowplow attack yeah okay
23:07Thomas don't get me started on autonomous robot arms attack I don't know if we can talk about this and
23:12Eric hold your horses we're not onto robots yet
23:16Thomas uh but this I actually think is like it's obvious so autumn so the my takeaway here is the fall a morning flight in the fall is my best chance to have an on-time flight right
23:30Eric exactly why weren't you traveling to the world cup in the shoulder seasons
23:35Thomas yeah I it's I mean I should have called infantino and told him to hold it
23:41Eric yeah
23:41September
23:42Thomas and October yeah
23:43Eric so
23:44Thomas the trends honestly this to me is the biggest takeaway like data backed and we all knew it but this proves the case fall mornings
23:54Eric the
23:55Thomas best okay
23:56Eric yeah and actually like every morning literally the mornings that's
24:01Thomas the biggest factor
24:02Eric they're nailing it
24:03Thomas if a mountain's if an avalanche comes down a mountain a chalet stops isn't moved by it and then you could leave from it first when
24:16Eric I was at school I studied cascading power failures um so think the big power outages of the northeast like 2003 there's been several some countries experience these on a regular basis rolling
24:29Thomas blackouts
24:29Eric rolling blackouts exactly exactly um and why they're interesting and maybe somewhat similar to this but some differences is um like in the power failure ones there's these rare events but when they happen they're really big and they're really big because one or two outages will weaken the system and then make it more likely for our future outages to happen right um in this case we have cascades it's a little different but it might be similar you know um so if we think about a delayed flight um that is gonna affect future flights and it's actually gonna change all this scheduling that a ton of people put a lot of time into right um that like single delayed flight out of some random airport into like Chicago you know like Chicago's got a lot of capacity so it might not affect it as much but you know if you get a few of these and they build up all of these queues that are in the airport start to become overloaded in various ways uh and then kind of uh they continue to delay they're weakening the system right all that capacity that we thought we had we now have more flights that we need to do because they're like carried over from the last hour and stuff right so they start over not a
25:48Thomas free-flowing channel there's build up god and i'm we're not there's gates there's the runways there's the air traffic coordinators to actually be able to like you know we're so many bottlenecks that yeah if it's a totally free-flowing machine sure everyone's gonna go in but yeah if you get some build up in one area of the system the downstream effect aka cascade yeah
26:12okay
26:13Eric so here's the math concepts if there's any math concepts to take away this is what i want so um why
26:20thomas's misery
26:22Eric trying to understand thomas's misery he chose his misery is mostly based that he chose to take a summer flight in the evening but uh consider that why are summer evenings so bad like why is that the case um so and this concept right here is an important one that you know i've been wanting to talk about a lot but let's get into it um so here we got some math symbols uh this time part
26:48i'm
26:49Thomas scared just i want you to know i'm now scared but we're gonna get through this but i'm scared
26:54Eric good try to keep it together for a little bit we've only got a few symbols to get through uh so the top think of as your arrival right so we're gonna simplify this airport you mentioned all these different complexities of an airport i'm gonna simplify it just imagine it as a single queue yeah we're not even gonna do multi-worker queue we just have a single queue so we have airplanes arriving queuing up and then we have airplanes leaving right so in this row what we call this is like the utilization of the system so it's how well it's being utilized um you know it's kind of overloaded you know it's like utilization is not necessarily good or bad and i'll get to that point
27:37Thomas um
27:38Eric but uh it's kind of yeah how well you're utilizing that right so if you had an exact match it would be great and there's no uncertainty
27:46Thomas there's
27:47Eric no uncertainty but uh almost all these things are thought of in probabilistic sense so we have some sort of arrival right but then it happens kind of randomly right and this happens in airports right so and in fast food joints you know you can expect 20 people an hour but when do they come and often you know they'll come in waves and stuff like that right and it's just kind of these oddities of statistics and probability and stuff there's a whole bunch of cool like thoughts around this uh these queuing systems
28:16Thomas uh actual arrival rate over capacity there's no perfect system i assume there's some typical number of like 70 percent is an ideal airport major airport in the united states rate what's the p what's the term we're using here
28:32Eric yeah so row utilization um there's other words for it i don't know that's the one that i often have used
28:40Thomas okay
28:40Eric but um this next point gets into what you're trying to talk about so if we assume some
28:47Thomas tries
28:48Eric that
28:48Thomas trying to trying to talk i
28:51Eric mean doing a good job because yeah it's like okay what do we do with this right why is it important because it's not obvious it's important the business owner probably doesn't care yet um but this next point brings it back to the real world so if you assume some kind of common probability distributions for arrival rate and throughput you know i'm just doing napkin math here i'm trying to you know just show some basic ways this operate so if you move your utilization to kind of a hundred percent what ends up happening is you're once you go over a hundred percent let's just think about that so let's assume our uh arrival rate is twice our throughput right what's going to happen to your queue
29:34Thomas full breakdown not zero right
29:38Eric the queue just keeps expanding right
29:39Thomas okay yeah
29:40Eric yeah so we're just gonna so like 10 people come i work through five 10 more people come
29:46got it yeah
29:46Thomas yeah it's like a
29:47Eric yeah
29:47infinite
29:48Eric so the queue actually keeps increasing if you ran the system forever your wait time will go to infinity right um so that tipping point is one 100 percent so once you get over 100 you can no longer keep up with the inbound traffic right like your weight you'll never get through your queue right but
30:06Thomas interestingly thought interesting thought experiment you still would land right like a plane would still land
30:13Eric oh yeah so this is more from the perspective of a single airport so the idea is all the planes would be landing half of them would be able to take off half couldn't so the amount of airports airplanes on the ground would just like keep growing
30:29Thomas not yeah so not all planes would take off it's not
30:33Eric yeah or
30:34Thomas they would take
30:35away they
30:36Eric would just keep growing yeah so you know this is kind of in theory so what would happen in the real world is they might have to divert aircrafts they would be circling or something you know at some point they run out of room on the ground to store all these inboard inbound airport airplanes yeah
30:51Thomas yeah because the buildup yeah because there's a physical space and so here's a question i don't understand why it wouldn't just be it wouldn't be like a parabolic curve because the full capacity means if the intake is equal to its capacity shouldn't everything be flowing like a river fully smooth and there's actually no delay so why is it when you're at 0.95 that you're starting to get delays wouldn't it only be once you're over capacity you're not flowing you know what i mean
31:27Eric yeah so if you had a system with no uncertainty it could potentially be like that but a lot of these are based on like a poisson arrival process which is random and what ends up happening is so say you expect 10 flights over the hour you don't actually get them every six minutes right you kind of end up like with waves and then the wave that comes you can only work through them you can't work through them probabilistically right they always take like six ten twenty minutes right to turn over so you don't kind of get that benefit on the outbound so what ends up happening is like on the inbound they come in this like chunky inbound and then they get queued up right
32:12Thomas okay yeah yeah so it's a similar cascading so this is another term it's the chunkiness
32:17also
32:19Thomas driver where it's not a smooth like of course if it was every five minutes everyone was coming in then you could flow like a river but it's chunked up so yeah okay
32:28Eric i mean this is something that all these businesses deal with you know it's like all these businesses have these like wavy inbound patterns and stuff especially like food places you know it's just natural with rhythms of you know sales
32:41Thomas cycles like this it's like such a yeah okay you
32:45Eric know um and then finally you know we're just like trying to understand you know everyone's misery here so it's like at the end of the day what ends up mattering is like how many passengers and how delayed they are right um and the whole thing the whole thing is like you can see the natural tension between everyone right so the air air aircraft carriers or yeah like the companies they're trying to drive row to the right utilization to the right
33:15Thomas you have more flights it's more money right
33:18Eric people want to drive it to the left they don't want to work
33:21Thomas zero delays
33:22Eric yep yep so what ends up happening is these companies is like all this like system kind of like uh self-organizes into an equilibrium right so what ends up happening is like companies might reduce slack and by reducing slack you're increasing utilization so that's that slack i keep talking about yeah like if you add slack you're basically uh moving it to the left if you're removing it you move it to the right yeah and when you do that the company if it works well they're gonna make more profit so they're like
33:59Thomas what's so interesting it's profit or misery
34:03Eric yeah and then what ends up happening they go a little too far yeah and the misery overwhelms and they're like screw this southwest sucks i'm never flying on them again right and then southwest has to change their practices to move it back the other way and that's why i describe it you know it's like this self-organizing equilibrium because you have all these different companies competing for these people the companies want to move right on this the people want it to move left and it finds the equilibrium uh and that trends thing is why it's interesting you know you kind of see that that uh the average delay time is directly i think related to this utilization factor and that's why i kind of made those like uh kind of like hand wavy statements at the beginning i'm like well i don't actually think because in the gut you know you're like um you feel like there's more slack time in the system right
34:56Thomas yeah yeah
34:57Eric but if that was true we should see average delays being less than they were in the past but they're not
35:04Thomas art well what about volume aren't there more flights now which can impact capacity so
35:10Eric glad you brought that up
35:11Thomas great uh
35:14Eric so that was the first thing i was thinking about too and you see it's basically flat which is shopping in its own way or
35:20Thomas um
35:20Eric i think that's it you know
35:22Thomas ofo
35:23Eric yeah the throughput's going left right um this is that's
35:27Thomas great i just keep thinking of a flowing river like beautiful beautiful nature always ooh nature almost always finds a one over one um throughput ratio be it flooding be it whatever gravity's gonna pull that down and it's gonna find its pathway and it's gonna get
35:55Eric yeah you're talking about like laws of physics and often when we do these simulations we create these constraints that are very much one-to-one right like each water molecule can't disappear right so every node it goes through it's got to come out of and stuff like that right and in this airport simulation there's a there's a lot of those so we've got a few other uh scenarios we don't have to go through them here's an example that's kind of cool it's a ground stop so at this point all the airplanes have stopped taking off these ones that were already in flight but you see there's almost an hour of no flights taking off and then check it out when they start taking off the whole system oh
36:37Thomas because they're all delayed damn
36:38Eric delayed yeah yeah is
36:40Thomas that a weather is that a weather event and
36:42Eric then look at these stats oh my god 38 on time it was a ground stop so this was some weird air traffic controller that's why it's so precise like it's literally an hour across the whole country
36:53Thomas can we say day or time
36:55Eric 2023 01 11
36:57Thomas january 11 2023 interesting
37:00Eric yeah what i was going was an what if machine to try to understand the incomprehensible variability um so i was like what is driving all this can we understand it can we see how things would propagate if we kind of cause events and do what ifs right so i've created this uh what if machine he has then i have this ignore baseline delay so you see how things are getting red and stuff
37:25Thomas yeah
37:26Eric so if i click this button everything's wait so we're just assuming all of the existing delays are expected yeah
37:32Thomas we don't care about them
37:34Eric yeah we're
37:35just trying
37:35Eric to understand what happens if we perturb the system a little yeah i
37:39got you
37:39Thomas i'm with you
37:40Eric all right and now we turn on god weather mode so click
37:44Thomas eric isn't happening right literally and
37:47Eric then now we start to contemplate what would be an interesting scenario so the
37:52Thomas ones up
37:53Eric so since you were flying out of chicago let's go early in the day because we know that kind of cascades throughout the day and that's what we kind of want to see right um so here east coast is taking off so right here we're gonna add some storms
38:09g this
38:12Thomas is like sim sim earth oh wow
38:15Eric so and now you see all these delayed flights taking off
38:19Thomas yeah out to the rest of the system and
38:20Eric they should start effect see denver's under stress now kind of absorbed it since it's such a big airport and there's only so much flights in between them but
38:31Thomas also you chicago's a united hub as is denver as is san francisco
38:35Eric yeah you're right that's probably part of it yeah there's more enter airport traffic
38:40Thomas yeah yeah
38:41Eric good point i like that
38:44Thomas so you just like uh severe thunderstorm in the morning and now yeah but
38:48Eric is that common does that actually happen in chicago morning thunder morning
38:52Thomas no to be honest no i mean you know like growing up minnesota it's thunderstorms more often in the evening
38:56Eric all right let's restart um i feel like most storms happen in the evening
39:02Thomas dude eric this is so awesome and impressive do
39:07Eric you feel like you learned some math today
39:09Thomas i do actually very much so the the throughput equation applies to basically everything especially in like any business that's serving customers or deals or whatever and the question is are you never going to be one over one and then so the question is what's the ideal flow rate where you can manage it and do it all effectively and successfully yeah
39:39Eric it's such a balancing act and there's a whole lot of heuristics around it you know like if you think of target you know they have multi-worker queues and stuff right and um you'll like if you show up and there's like four people in line it's like a miserable experience right um but you got to understand to get it to not have four people in line at random parts throughout the day you have to have like non-trivially lower utilization than 100 right you gotta be down at like 80 or something or you have to have like slack that comes from else
40:16Thomas i think we can now tally the score so i think it's not i think there's vibes in play but i think there you produce substantial evidence to show that there is more math and modeling than vibes yeah math wins um to be honest the biggest nugget i have because it impacts daily life is the cast the risk of cascading impact and morning floods which i think people generally know but it's like so apparent yeah so yeah but i the utilization curve and trying to hit it but that's more fucking you know in the air you
40:59Eric know
40:59unintended
41:00Thomas like that i think there's multiple nuggets but i think the seasonality was interesting but really it's and it's so obvious it's like get on the first flight out of the airport and you're gonna be on time
41:10Eric yeah you can't have these cascades
41:14Thomas yeah mark market market dude can you
41:18Eric win it
41:18Thomas yeah math
41:20Eric we're locked in
41:21Thomas math wins this round for the first time ever in the history of the world