WEBVTT 00:00:00.000 --> 00:00:02.980 Welcome in. It is really fantastic to have you 00:00:02.980 --> 00:00:04.839 here with us today. Yeah, it's great to be here. 00:00:05.000 --> 00:00:07.639 Whether you are prepping for a major presentation 00:00:07.639 --> 00:00:12.160 or trying to catch up on a totally new field, 00:00:12.259 --> 00:00:16.039 or honestly, if you are just insanely curious 00:00:16.039 --> 00:00:18.660 about how the underlying systems of the world 00:00:18.660 --> 00:00:21.679 actually work. You have come to the exact right 00:00:21.679 --> 00:00:23.960 place. The absolute best place. We have a highly 00:00:23.960 --> 00:00:26.500 focused mission for this deep dive. Today we 00:00:26.500 --> 00:00:28.660 are exploring the really fascinating and sometimes 00:00:28.660 --> 00:00:31.699 hidden intersection of advanced mathematics and 00:00:31.699 --> 00:00:33.579 men's college basketball. Which is quite the 00:00:33.579 --> 00:00:35.920 combination. It is. And we are going to do it 00:00:35.920 --> 00:00:38.320 by extracting the most crucial insights from 00:00:38.320 --> 00:00:41.210 a single source. which is a comprehensive breakdown 00:00:41.210 --> 00:00:43.590 of the Pomeroy College basketball ratings. And 00:00:43.590 --> 00:00:45.469 I actually love the visual we have going on right 00:00:45.469 --> 00:00:47.030 behind you. Well, yeah, the glowing basketball 00:00:47.030 --> 00:00:49.210 court. Yeah, with all those complex mathematical 00:00:49.210 --> 00:00:51.549 equations and scrolling statistics overlaid on 00:00:51.549 --> 00:00:53.950 the hardwood, it perfectly sets the mood. I am 00:00:53.950 --> 00:00:56.270 very ready to dig into this with you because 00:00:56.270 --> 00:00:59.509 we are looking at a space where the visceral, 00:00:59.509 --> 00:01:02.649 highly emotional reality of a basketball court 00:01:02.649 --> 00:01:06.230 collides directly with pure abstract mathematics. 00:01:07.170 --> 00:01:09.879 It's a perfect study in how data can completely 00:01:09.879 --> 00:01:12.840 upend traditional wisdom. Absolutely. And before 00:01:12.840 --> 00:01:14.700 we jump in, I just want to set the table for 00:01:14.700 --> 00:01:16.859 you. This is a deep dive. We are taking this 00:01:16.859 --> 00:01:19.459 specific source material, stripping away the 00:01:19.459 --> 00:01:22.060 noise, and pulling out the absolute most important 00:01:22.060 --> 00:01:24.780 concepts for you. Just the essentials. Exactly. 00:01:24.840 --> 00:01:27.480 We are looking at how a niche mathematical model 00:01:27.480 --> 00:01:30.939 revolutionized the way analysts, media heavyweights, 00:01:30.939 --> 00:01:33.280 and fans understand one of the most unpredictable 00:01:33.280 --> 00:01:36.500 sports in America. Okay, let's unpack this. Let's 00:01:36.500 --> 00:01:38.319 do it. We have to start at the very beginning 00:01:38.319 --> 00:01:41.180 with the origins of the system. The ratings are 00:01:41.180 --> 00:01:43.719 essentially a series of predictive models for 00:01:43.719 --> 00:01:46.519 men's college basketball teams created by Ken 00:01:46.519 --> 00:01:49.560 Pomeroy. But the detail that really stands out 00:01:49.560 --> 00:01:52.000 to me is the timeline. The timeline is key. These 00:01:52.000 --> 00:01:54.079 ratings were first published online in the year 00:01:54.079 --> 00:01:56.959 2003, and they were completely free of charge. 00:01:57.200 --> 00:01:58.939 Think about the internet and sports analysis 00:01:58.939 --> 00:02:01.659 back in 2003. We were still dealing with dial 00:02:01.659 --> 00:02:05.219 -up in a lot of places. Yeah, that 2003 timestamp 00:02:05.219 --> 00:02:08.400 is vital context. If you think about the landscape 00:02:08.400 --> 00:02:10.960 back then, we were definitely not saturated with 00:02:10.960 --> 00:02:13.460 the kind of advanced metrics we see today. The 00:02:13.460 --> 00:02:16.699 analysis of the era was heavily, heavily reliant 00:02:16.699 --> 00:02:19.340 on the eye test. The eye test? Literally just 00:02:19.340 --> 00:02:22.620 what a scout or a coach or a columnist felt when 00:02:22.620 --> 00:02:25.360 they watched a team play. It was dominated by 00:02:25.360 --> 00:02:27.300 subjective narrative. Right. It was all about 00:02:27.300 --> 00:02:29.419 momentum and heart and... Who wants it more? 00:02:29.580 --> 00:02:32.139 Exactly. Those classic who wants it more arguments. 00:02:32.360 --> 00:02:34.599 You'd hear on sports talk radio constantly. The 00:02:34.599 --> 00:02:37.199 primary data points people relied on were simple 00:02:37.199 --> 00:02:39.319 box scores, you know, points, rebounds, assists, 00:02:39.560 --> 00:02:42.180 and obviously a team's overall win loss record. 00:02:42.240 --> 00:02:46.539 Right. So into that very specific. somewhat archaic 00:02:46.539 --> 00:02:49.900 environment steps, Ken Pomeroy publishing a highly 00:02:49.900 --> 00:02:52.560 sophisticated predictive mathematical model online 00:02:52.560 --> 00:02:56.060 for anyone to access. It was essentially democratizing 00:02:56.060 --> 00:02:58.500 high -level sports data before that was really 00:02:58.500 --> 00:03:00.340 a thing. It was a massive leap forward because 00:03:00.340 --> 00:03:02.580 what he was publishing wasn't just a slightly 00:03:02.580 --> 00:03:05.360 better way of counting basic stats. He was introducing 00:03:05.870 --> 00:03:08.129 a fundamentally different way to evaluate performance. 00:03:08.509 --> 00:03:11.150 Well, the core mathematical engine driving the 00:03:11.150 --> 00:03:13.710 entire system is based on something called the 00:03:13.710 --> 00:03:17.409 Pythagorean expectation, which Pomeroy specifically 00:03:17.409 --> 00:03:21.710 adjusted for the unique nuances of college basketball. 00:03:21.930 --> 00:03:23.569 Now I have to play devil's advocate here for 00:03:23.569 --> 00:03:25.669 a second. Go for it. If you were following a 00:03:25.669 --> 00:03:28.090 college basketball season, the ultimate goal 00:03:28.090 --> 00:03:32.150 is to win games. If a team is 25 and 5, they 00:03:32.150 --> 00:03:36.050 are winning. Why do we need a complex Pythagorean 00:03:36.050 --> 00:03:39.569 expectation to tell us how good they are? Isn't 00:03:39.569 --> 00:03:41.430 the win -loss record the ultimate indicator? 00:03:41.849 --> 00:03:44.210 What's fascinating here is that the math actually 00:03:44.210 --> 00:03:47.009 proves a win -loss record can be incredibly deceptive. 00:03:47.069 --> 00:03:49.629 Really? Yeah. If we look at the concepts tied 00:03:49.629 --> 00:03:52.189 to this model, specifically terms from the source 00:03:52.189 --> 00:03:55.870 like sabermetrics and win probability, we see 00:03:55.870 --> 00:03:58.169 a total paradigm shift. Traditional evaluation 00:03:58.169 --> 00:04:01.050 just looked at the final score, but Saber metrics 00:04:01.050 --> 00:04:03.009 and the Pythagorean expectation look under the 00:04:03.009 --> 00:04:04.870 hood. They look at the engine. Exactly. They 00:04:04.870 --> 00:04:07.110 evaluate the underlying metrics. Specifically, 00:04:07.250 --> 00:04:09.270 they look at the ratio of points a team scores 00:04:09.270 --> 00:04:11.430 compared to the points they allow over time. 00:04:11.789 --> 00:04:13.569 Okay, so point differential. But why is that 00:04:13.569 --> 00:04:15.330 inherently better than just looking at wins? 00:04:15.590 --> 00:04:17.949 Because a simple win -loss record doesn't account 00:04:17.949 --> 00:04:21.269 for luck or variance or the actual margin of 00:04:21.269 --> 00:04:24.230 victory. Imagine a team that wins five games 00:04:24.230 --> 00:04:27.410 in a row by a single point. maybe off some lucky 00:04:27.410 --> 00:04:29.769 buzzer beaters. Sure. Their record looks flawless, 00:04:30.129 --> 00:04:32.529 but the math looks at their point differential 00:04:32.529 --> 00:04:35.610 and realizes they are living on a knife's edge. 00:04:35.829 --> 00:04:38.470 Now compare them to a team that loses one game 00:04:38.470 --> 00:04:41.149 by two points, but wins their next four games 00:04:41.149 --> 00:04:43.550 by 30 points each. Ah, I see where you're going 00:04:43.550 --> 00:04:46.189 with this. Right. The second team has a technically 00:04:46.189 --> 00:04:49.370 worse win -loss record, but the Pythagorean expectation 00:04:49.370 --> 00:04:52.649 reveals that they are vastly secure in their 00:04:52.649 --> 00:04:54.829 possession -by -possession efficiency. So it 00:04:54.829 --> 00:04:58.000 exposes the paper tigers of the sport. It moves 00:04:58.000 --> 00:05:00.180 the conversation away from a historical record 00:05:00.180 --> 00:05:02.879 of what happened yesterday and transforms it 00:05:02.879 --> 00:05:05.079 into a predictive engine calculating what is 00:05:05.079 --> 00:05:07.579 statistically probable to happen tomorrow. Precisely. 00:05:07.579 --> 00:05:09.600 And the math itself isn't just a simple linear 00:05:09.600 --> 00:05:12.899 ratio. The Pythagorean expectation applies exponents 00:05:12.899 --> 00:05:15.160 to those points scored and points allowed. Hence 00:05:15.160 --> 00:05:17.759 the Pythagorean name. Right, mimicking the famous 00:05:17.759 --> 00:05:20.220 formula. In baseball, where this originated, 00:05:20.519 --> 00:05:38.170 the exponent was roughly two. And it is really 00:05:38.170 --> 00:05:40.899 worth noting that Our source makes it clear Pomeroy 00:05:40.899 --> 00:05:43.720 wasn't operating in a total vacuum. He is part 00:05:43.720 --> 00:05:46.620 of a very specific lineage of thinkers. He is. 00:05:46.680 --> 00:05:49.339 The text names two other noted statisticians 00:05:49.339 --> 00:05:51.620 who use variations of this exact mathematical 00:05:51.620 --> 00:05:55.060 concept in basketball, Dean Oliver and John Hollinger. 00:05:55.279 --> 00:05:57.899 When you see a creator like Pomeroy linked to 00:05:57.899 --> 00:06:00.620 figures like Oliver and Hollinger, it establishes 00:06:00.620 --> 00:06:02.879 the credibility of the whole movement. These 00:06:02.879 --> 00:06:05.160 are the foundational architects of modern basketball 00:06:05.160 --> 00:06:08.000 analytics. Pomeroy is taking the analytical frameworks 00:06:08.000 --> 00:06:10.209 that were Just starting to gain traction in professional 00:06:10.209 --> 00:06:12.290 sports and custom building an engine for the 00:06:12.290 --> 00:06:14.709 wildly unpredictable world of college hoops. 00:06:14.790 --> 00:06:17.750 Custom building that engine is key because college 00:06:17.750 --> 00:06:20.970 basketball isn't the NBA. You don't have 30 teams 00:06:20.970 --> 00:06:23.310 playing a standardized 82 -game schedule. Far 00:06:23.310 --> 00:06:27.649 from it. You have hundreds of Division I programs, 00:06:27.970 --> 00:06:30.470 wildly different levels of talent, and non -conference 00:06:30.470 --> 00:06:33.230 schedules that are just all over the map. You 00:06:33.230 --> 00:06:36.009 can have all the elegant math in the world. But 00:06:36.009 --> 00:06:38.370 if your model doesn't accurately predict the 00:06:38.370 --> 00:06:41.769 games in that chaotic environment, it is completely 00:06:41.769 --> 00:06:44.410 useless. Exactly. The proof of concept always 00:06:44.410 --> 00:06:46.170 has to be in the prediction. Here's where it 00:06:46.170 --> 00:06:48.269 gets really interesting. Our source references 00:06:48.269 --> 00:06:51.350 a 2011 article from the New York Times, specifically 00:06:51.350 --> 00:06:54.110 from Nate Silver's FiveThirtyEight blog, which, 00:06:54.250 --> 00:06:56.689 as you likely know, is an outlet famously dedicated 00:06:56.689 --> 00:07:00.290 to rigorous data -driven journalism. Oh, absolutely. 00:07:00.889 --> 00:07:02.910 In that piece, they reveal a hard data point 00:07:02.910 --> 00:07:05.310 that is really the absolute anchor of this deep 00:07:05.310 --> 00:07:07.949 dive. According to their analysis in 2011, the 00:07:07.949 --> 00:07:10.290 Pomeroy College basketball readings boasted a 00:07:10.290 --> 00:07:13.519 73 % success rate in their predictions. That 00:07:13.519 --> 00:07:16.779 73 % figure is phenomenal, especially when you 00:07:16.779 --> 00:07:19.339 consider the sheer volume and volatility of college 00:07:19.339 --> 00:07:22.399 basketball games. It is. But the source immediately 00:07:22.399 --> 00:07:24.680 gives us a point of comparison to put that 73 00:07:24.680 --> 00:07:27.360 % into perspective. It notes that this success 00:07:27.360 --> 00:07:30.759 rate is exactly 2 % better than the ratings percentage 00:07:30.759 --> 00:07:33.759 index commonly known as the RPI. Right. Now, 00:07:33.779 --> 00:07:35.839 for a long time, the RPI was the gold standard 00:07:35.839 --> 00:07:38.939 metric used by the establishment, including the 00:07:38.939 --> 00:07:41.399 actual tournament selection committee, to evaluate 00:07:41.399 --> 00:07:44.720 teams. Yes. RPI is explicitly listed in the source 00:07:44.720 --> 00:07:47.019 under traditional methods and computer models. 00:07:47.180 --> 00:07:50.000 It was the entrenched system. So Pomeroy's open 00:07:50.000 --> 00:07:52.399 source math beats the establishment's RPI by 00:07:52.399 --> 00:07:55.319 2%. I have to push back again here, though. OK. 00:07:55.379 --> 00:07:58.779 If you were just glancing at those numbers, 73 00:07:58.779 --> 00:08:01.639 % versus 71%, a 2 % margin sounds incredibly 00:08:01.639 --> 00:08:04.300 small. You might look at that and think, we are 00:08:04.300 --> 00:08:06.560 doing all this complex possession by possession 00:08:06.560 --> 00:08:09.699 Pythagorean math, applying exponents to point 00:08:09.699 --> 00:08:13.040 ratios just to gain two measly percentage points 00:08:13.040 --> 00:08:15.279 over the traditional formula. It is a totally 00:08:15.279 --> 00:08:17.839 fair question. But in this specific context, 00:08:18.100 --> 00:08:20.939 a 2 % edge is not small at all. It is a monumental 00:08:20.939 --> 00:08:23.839 advantage. To understand why, you have to look 00:08:23.839 --> 00:08:26.649 at the concepts listed in the text. specifically 00:08:26.649 --> 00:08:29.050 strength of schedule and home advantage. And 00:08:29.050 --> 00:08:31.230 you have to compare how Pomeroy handles them 00:08:31.230 --> 00:08:33.830 versus how the RPI handled them. Let's break 00:08:33.830 --> 00:08:36.450 that down. How did the RPI handle it? The RPI 00:08:36.450 --> 00:08:39.529 was heavily reliant on just three factors. A 00:08:39.529 --> 00:08:41.590 team's winning percentage, their opponent's winning 00:08:41.590 --> 00:08:43.750 percentage, and their opponent's opponent's winning 00:08:43.750 --> 00:08:45.710 percentage. So it was basically a giant strength 00:08:45.710 --> 00:08:48.899 of schedule calculator. Exactly. But... The fatal 00:08:48.899 --> 00:08:51.559 flaw was that it didn't care how you played against 00:08:51.559 --> 00:08:53.740 that schedule. If you played a team with a great 00:08:53.740 --> 00:08:57.200 record and lost by 40 points, the RPI might actually 00:08:57.200 --> 00:08:59.279 reward you just because your opponent was highly 00:08:59.279 --> 00:09:02.320 rated. It was totally blind to efficiency. That 00:09:02.320 --> 00:09:05.200 sounds like a massive blind spot. It was. Pomeroy's 00:09:05.200 --> 00:09:07.919 adjusted Pythagorean system corrects this entirely. 00:09:08.320 --> 00:09:11.340 His model evaluates efficiency on a per possession 00:09:11.340 --> 00:09:14.379 basis. It strips away the pace of the game. So 00:09:14.379 --> 00:09:17.720 if a team plays a very slow, deliberate style 00:09:17.720 --> 00:09:20.159 and only scores 60 points, but they only had 00:09:20.159 --> 00:09:24.139 50 possessions, Pomeroy's math recognizes that 00:09:24.139 --> 00:09:27.049 they are incredibly efficient. It then takes 00:09:27.049 --> 00:09:29.909 that efficiency and adjusts it based on the defensive 00:09:29.909 --> 00:09:31.970 strength of the specific opponent they played. 00:09:32.230 --> 00:09:34.370 And it mathematically accounts for whether the 00:09:34.370 --> 00:09:37.090 game was played at home or on the road. So how 00:09:37.090 --> 00:09:39.570 does that directly translate to the 2 % advantage? 00:09:39.970 --> 00:09:42.629 Think about it like the house edge in a casino. 00:09:43.230 --> 00:09:45.929 A casino doesn't need to win 90 % of the hands 00:09:45.929 --> 00:09:48.169 at a blackjack table to be wildly profitable. 00:09:48.389 --> 00:09:50.850 They just need a mathematical edge of a couple 00:09:50.850 --> 00:09:52.610 of percentage points over the player. Right, 00:09:52.690 --> 00:09:55.409 because of the volume. Yes. Over the course of 00:09:55.409 --> 00:09:57.850 thousands of hands, that tiny edge translates 00:09:57.850 --> 00:10:00.850 into an insurmountable advantage. The exact same 00:10:00.850 --> 00:10:03.210 principle applies here. Over the course of a 00:10:03.210 --> 00:10:05.409 massive college basketball season, with literally 00:10:05.409 --> 00:10:08.149 thousands of individual games being played consistently 00:10:08.149 --> 00:10:11.370 beating the established metric by 2%, means your 00:10:11.370 --> 00:10:13.909 underlying model is capturing reality far more 00:10:13.909 --> 00:10:16.409 accurately than anything else out there. It means 00:10:16.409 --> 00:10:19.049 your math is successfully quantifying the impact 00:10:19.049 --> 00:10:22.350 of a brutal road environment or the true efficiency 00:10:22.350 --> 00:10:25.610 of a slow -paced team in a way that traditional 00:10:25.610 --> 00:10:30.029 RPI was completely missing. Yes. That 2 % edge 00:10:30.029 --> 00:10:32.649 is the mathematical proof that Pomeroy's adjustments 00:10:32.649 --> 00:10:35.029 for pace schedule and home court are working. 00:10:35.389 --> 00:10:37.769 It is why data journalists at the New York Times 00:10:37.769 --> 00:10:40.649 were paying such close attention. He built a 00:10:40.649 --> 00:10:42.950 better mousetrap. And the source groups these 00:10:42.950 --> 00:10:44.929 rating systems into different categories, which 00:10:44.929 --> 00:10:47.669 sets up a really fascinating contrast in how 00:10:47.669 --> 00:10:50.100 we evaluate sports as a whole. If we connect 00:10:50.100 --> 00:10:52.080 this to the bigger picture, we see a fundamental 00:10:52.080 --> 00:10:55.000 clash of evaluation philosophies detailed right 00:10:55.000 --> 00:10:57.500 here in the text. Under the sports rating systems 00:10:57.500 --> 00:11:00.159 banner, we see Pomeroy listed alongside other 00:11:00.159 --> 00:11:02.759 objective mathematical models. But right beneath 00:11:02.759 --> 00:11:04.820 that, there is an entirely different category 00:11:04.820 --> 00:11:07.820 labeled polls and opinion. This includes the 00:11:07.820 --> 00:11:09.639 traditional heavyweights of sports media like 00:11:09.639 --> 00:11:12.259 the AP poll and the coaches poll. Ah, yes, the 00:11:12.259 --> 00:11:14.879 classic top 25 rankings. The lists that drive 00:11:14.879 --> 00:11:17.379 90 % of the debates are on the water cooler and 00:11:17.379 --> 00:11:20.149 on sports television. Exactly. And it is crucial 00:11:20.149 --> 00:11:22.450 to recognize that we aren't saying one is inherently 00:11:22.450 --> 00:11:24.529 good and the other is bad. We are simply looking 00:11:24.529 --> 00:11:26.529 at the factual difference between an objective 00:11:26.529 --> 00:11:29.490 algorithm and subjective human voting. Which 00:11:29.490 --> 00:11:32.629 is vast. Very. The Pomeroy ratings represent 00:11:32.629 --> 00:11:36.450 a cold, emotionless mathematical reality. The 00:11:36.450 --> 00:11:39.029 algorithm does not care if a team has a legendary 00:11:39.029 --> 00:11:41.870 Hall of Fame coach. It doesn't care if a mid 00:11:41.870 --> 00:11:44.769 -major program has an incredibly compelling underdog 00:11:44.769 --> 00:11:47.590 storyline that everyone is rooting for. It definitely 00:11:47.590 --> 00:11:50.190 doesn't - care if a player hit a flashy highlight 00:11:50.190 --> 00:11:52.649 reel dunk on national television last night? 00:11:52.850 --> 00:11:55.230 It is entirely blind to the narrative. It just 00:11:55.230 --> 00:11:57.710 sees the data points. Precisely. It only cares 00:11:57.710 --> 00:11:59.850 about the raw data of possessions, offensive 00:11:59.850 --> 00:12:02.450 efficiency and defensive efficiency. On the other 00:12:02.450 --> 00:12:04.590 hand, the subjective human voting systems like 00:12:04.590 --> 00:12:06.730 the AP poll, which is driven by sports writers 00:12:06.730 --> 00:12:09.690 or the coaches poll, are entirely based on human 00:12:09.690 --> 00:12:12.049 observation. And human observation naturally 00:12:12.049 --> 00:12:14.870 incorporates narrative momentum and bias. For 00:12:14.870 --> 00:12:17.970 example, human voters are notorious. for penalizing 00:12:17.970 --> 00:12:20.669 a top -ranked team heavily if they barely survive 00:12:20.669 --> 00:12:23.029 an ugly low -scoring game against an unranked 00:12:23.029 --> 00:12:25.570 opponent. Even if they win, the voters drop them 00:12:25.570 --> 00:12:28.789 in the polls. Or conversely, human voters might 00:12:28.789 --> 00:12:31.149 reward a team for showing incredible grit and 00:12:31.149 --> 00:12:34.529 heart in a close loss to a rival. Right. Humans 00:12:34.529 --> 00:12:37.120 judge the aesthetics of the game. A mathematical 00:12:37.120 --> 00:12:39.659 model just calculates the adjusted point differential. 00:12:40.240 --> 00:12:42.440 What is so valuable for you as someone trying 00:12:42.440 --> 00:12:44.580 to gain a deeper understanding of the sport is 00:12:44.580 --> 00:12:46.480 recognizing that holding multiple perspectives 00:12:46.480 --> 00:12:49.320 gives you a massive analytical edge. You need 00:12:49.320 --> 00:12:51.980 both. Yes. If you only look at the subjective 00:12:51.980 --> 00:12:54.279 human poles, you get swept up in the hype and 00:12:54.279 --> 00:12:56.820 miss the underlying inefficiencies. If you only 00:12:56.820 --> 00:12:59.480 look at the math, you might miss unquantifiable 00:12:59.480 --> 00:13:01.860 human elements like injuries or team chemistry 00:13:01.860 --> 00:13:04.100 issues that aren't yet reflected in the data. 00:13:04.259 --> 00:13:06.960 You use the math. to check your inherent human 00:13:06.960 --> 00:13:09.600 biases, and you use your human observation to 00:13:09.600 --> 00:13:11.700 contextualize the math. That makes total sense. 00:13:11.940 --> 00:13:14.279 And what is so incredible to see in this source 00:13:14.279 --> 00:13:18.340 material is how rapidly Pomeroy's cold mathematical 00:13:18.340 --> 00:13:21.820 approach bled into the mainstream sports media 00:13:21.820 --> 00:13:24.539 consciousness. We are talking about an environment 00:13:24.539 --> 00:13:26.980 that was deeply entrenched in subjective opinions, 00:13:27.120 --> 00:13:30.360 and the math just forced its way in. The citations 00:13:30.360 --> 00:13:32.440 in this Wikipedia article serve as a roadmap 00:13:32.440 --> 00:13:34.740 of how advanced analytics conquered traditional 00:13:34.740 --> 00:13:37.179 sports media. It didn't happen overnight, but 00:13:37.179 --> 00:13:47.059 the progression is undeniable. and specialized 00:13:47.059 --> 00:13:49.200 data -heavy sites like Basketball Perspectives. 00:13:49.220 --> 00:13:51.440 That makes perfect sense. That is his core demographic 00:13:51.440 --> 00:13:53.919 of early adopters. The stat heads. Right. But 00:13:53.919 --> 00:13:56.340 then the circle expands. You see citations from 00:13:56.340 --> 00:13:58.899 ESPN's College Basketball Nation blog. Then it 00:13:58.899 --> 00:14:01.299 jumps to mainstream print media, both local and 00:14:01.299 --> 00:14:04.019 national. The source lists pieces from the Topeka 00:14:04.019 --> 00:14:06.080 Capital Journal all the way to the New York Daily 00:14:06.080 --> 00:14:08.899 News. There's even a specific citation from 2010 00:14:08.899 --> 00:14:11.000 where the Daily News discusses Dick Fetella, 00:14:11.200 --> 00:14:14.779 who is arguably the ultimate embodiment of traditional 00:14:14.779 --> 00:14:17.899 emotion -driven enthusiastic broadcasting in 00:14:17.899 --> 00:14:20.580 the exact same broader conversation about tournament 00:14:20.580 --> 00:14:24.340 selection alongside Pomeroy's metrics. Wow. The 00:14:24.340 --> 00:14:26.960 contrast between those two approaches being mentioned 00:14:26.960 --> 00:14:29.360 in the same breath is striking. It really is 00:14:29.360 --> 00:14:31.799 the old guard meeting the new guard. But for 00:14:31.799 --> 00:14:33.799 me, the ultimate flex in these citations are 00:14:33.799 --> 00:14:35.879 the articles from the Wall Street Journal. The 00:14:35.879 --> 00:14:37.899 source notes multiple pieces written by Carl 00:14:37.899 --> 00:14:39.860 Bielek for The Wall Street Journal's Daily Fix. 00:14:40.100 --> 00:14:43.580 That is a pivotal crossover moment. When a financial 00:14:43.580 --> 00:14:46.120 newspaper improved a publication whose entire 00:14:46.120 --> 00:14:48.700 reputation is built on rigorous data analysis, 00:14:48.799 --> 00:14:50.940 market efficiencies, and quantitative modeling, 00:14:51.120 --> 00:14:53.220 starts routinely citing your college basketball 00:14:53.220 --> 00:14:55.679 ratings, you have officially established supreme 00:14:55.679 --> 00:14:58.460 credibility. Look at the titles of those WSJ 00:14:58.460 --> 00:15:01.440 pieces provided in the references. Preseason 00:15:01.440 --> 00:15:04.220 rankings should be heated in Ken Pomeroy's winning 00:15:04.220 --> 00:15:06.980 bracket. They aren't treating this as a fun sports 00:15:06.980 --> 00:15:09.200 distraction. They are treating his basketball 00:15:09.200 --> 00:15:12.399 algorithm with the exact same seriousness and 00:15:12.399 --> 00:15:15.080 analytical rigor one might apply to a complex 00:15:15.080 --> 00:15:18.159 stock market forecast. Exactly. They recognize 00:15:18.159 --> 00:15:21.740 that a 73 percent predictive accuracy is a massive 00:15:21.740 --> 00:15:24.340 market inefficiency just waiting to be exploited. 00:15:24.559 --> 00:15:27.340 And that leads directly to the ultimate proof 00:15:27.340 --> 00:15:29.639 of his elevated status within the sports media 00:15:29.639 --> 00:15:32.379 ecosystem, which is explicitly noted right there 00:15:32.379 --> 00:15:34.769 in the text. Pomeroy became a contributing writer 00:15:34.769 --> 00:15:37.309 for ESPN's Insider Feature. Think about that 00:15:37.309 --> 00:15:39.169 arc for a second. We started this conversation 00:15:39.169 --> 00:15:41.909 talking about a solitary guy who published a 00:15:41.909 --> 00:15:44.350 free experimental mathematical model on a website 00:15:44.350 --> 00:15:47.409 back in 2003, right when people were still arguing 00:15:47.409 --> 00:15:49.690 over newspaper box scores. Over the course of 00:15:49.690 --> 00:15:52.549 the next decade, because that adjusted Pythagorean 00:15:52.549 --> 00:15:56.190 expectation proved to be so devastatingly accurate 00:15:56.190 --> 00:15:59.210 hitting that 2 % edge over the establishment 00:15:59.210 --> 00:16:03.279 RPI. It completely evolves. The free math becomes 00:16:03.279 --> 00:16:06.559 premium, highly sought after expertise locked 00:16:06.559 --> 00:16:08.980 behind the subscription paywall of the biggest 00:16:08.980 --> 00:16:11.759 sports broadcasting network on the planet. It 00:16:11.759 --> 00:16:14.320 perfectly illustrates a core truth about highly 00:16:14.320 --> 00:16:16.840 competitive industries. The establishment will 00:16:16.840 --> 00:16:19.200 always adopt a new paradigm once its predictive 00:16:19.200 --> 00:16:21.940 power is undeniably proven. The traditional media 00:16:21.940 --> 00:16:23.980 didn't just ignore the math, they absorbed it. 00:16:24.399 --> 00:16:26.220 They recognized that the Pythagorean expectation 00:16:26.220 --> 00:16:28.840 properly adjusted for strength of schedule and 00:16:28.840 --> 00:16:30.720 home court advantage on a per -possession basis 00:16:30.720 --> 00:16:34.340 was simply a better tool. The Nate Silver 538 00:16:34.340 --> 00:16:37.539 article titled, In NCAA Tournament Overachievers, 00:16:37.740 --> 00:16:40.879 Often Disappoint, really sums up exactly why 00:16:40.879 --> 00:16:43.659 the media needed this tool. The math exposes 00:16:43.659 --> 00:16:46.159 the overachievers. It strips away the illusion 00:16:46.159 --> 00:16:48.700 of a lucky five -game winning streak and reveals 00:16:48.700 --> 00:16:50.700 the true underlying strength or weakness of a 00:16:50.700 --> 00:16:53.200 team. Exactly. It gives analysts an x -ray vision 00:16:53.200 --> 00:16:55.580 that the eye test simply cannot provide. And 00:16:55.580 --> 00:16:58.519 once you have that x -ray vision, you can't unsee 00:16:58.519 --> 00:17:01.100 the underlying structure. It forces everyone 00:17:01.100 --> 00:17:03.539 from the casual fan filling out a bracket to 00:17:03.539 --> 00:17:05.859 professional analysts on television to elevate 00:17:05.859 --> 00:17:08.359 their arguments. You can no longer just say they 00:17:08.359 --> 00:17:10.859 wanted it more. You have to explain how they 00:17:10.859 --> 00:17:13.759 overcame a massive deficit in offensive rebounding 00:17:13.759 --> 00:17:17.049 efficiency. So what does this all mean? We have 00:17:17.049 --> 00:17:19.450 covered a vast amount of ground today, pulling 00:17:19.450 --> 00:17:22.470 apart the history, the math, and the impact from 00:17:22.470 --> 00:17:25.349 this single source. We took a journey that started 00:17:25.349 --> 00:17:28.670 in 2003 with a statistician deciding to apply 00:17:28.670 --> 00:17:32.069 an adjusted Pythagorean expectation. to the wildly 00:17:32.069 --> 00:17:34.109 unpredictable world of men's college basketball. 00:17:34.289 --> 00:17:36.529 We did. We explored how shifting the focus from 00:17:36.529 --> 00:17:39.210 simple wins and losses to a deeper per position 00:17:39.210 --> 00:17:41.890 efficiency metric allowed this model to outperform 00:17:41.890 --> 00:17:43.589 the traditional establishment backed ratings 00:17:43.589 --> 00:17:47.009 percentage index by a vital 2 % margin. And we 00:17:47.009 --> 00:17:49.349 broke down why that 2 % is actually a massive 00:17:49.349 --> 00:17:51.450 casino like edge in a high variance environment. 00:17:51.789 --> 00:17:56.130 Exactly. And we traced. How that undeniable 73 00:17:56.130 --> 00:17:59.230 % predictive accuracy ultimately forced its way 00:17:59.230 --> 00:18:01.970 into the mainstream, changing the way heavyweights 00:18:01.970 --> 00:18:04.930 from ESPN to the data journalists of the New 00:18:04.930 --> 00:18:06.430 York Times and the Wall Street Journal talk about 00:18:06.430 --> 00:18:08.230 analyze and understand the sport. It changed 00:18:08.230 --> 00:18:10.170 everything. For you, whether you are trying to 00:18:10.170 --> 00:18:12.990 find an edge in your office tournament bracket 00:18:12.990 --> 00:18:16.349 or you are just someone fascinated by how quantitative 00:18:16.349 --> 00:18:19.710 data shapes our understanding of reality, the 00:18:19.710 --> 00:18:22.069 story of the Pomeroy ratings offers a profound 00:18:22.069 --> 00:18:25.109 lesson. It proves that if you're willing to look 00:18:25.109 --> 00:18:28.109 past surface -level results, like a team's basic 00:18:28.109 --> 00:18:30.829 record or the loudest, most confident opinions 00:18:30.829 --> 00:18:33.349 on television, and dig into the underlying objective 00:18:33.349 --> 00:18:36.269 math, you can gain a measurable, distinct advantage. 00:18:36.789 --> 00:18:39.230 The truth of the situation is very often found 00:18:39.230 --> 00:18:41.329 in the deeper metrics hidden just beneath the 00:18:41.329 --> 00:18:43.450 popular narrative. This raises an important question, 00:18:43.490 --> 00:18:45.910 though. We have spent this entire time dissecting 00:18:45.910 --> 00:18:48.690 and marveling at a 73 % predictive success rate. 00:18:48.960 --> 00:18:51.700 And as we established, hitting that mark in an 00:18:51.700 --> 00:18:54.259 environment as notoriously chaotic, emotional 00:18:54.259 --> 00:18:57.220 and high variance as college basketball is a 00:18:57.220 --> 00:19:00.380 phenomenal mathematical achievement. But by definition, 00:19:00.539 --> 00:19:03.180 it still leaves 27 percent of the outcomes completely 00:19:03.180 --> 00:19:06.279 unpredicted by the algorithm. If models like 00:19:06.279 --> 00:19:08.579 the Pomeroy ratings can successfully strip away 00:19:08.579 --> 00:19:11.039 human bias and subjective narrative to achieve 00:19:11.039 --> 00:19:14.420 that 73 % accuracy today, will we eventually 00:19:14.420 --> 00:19:17.119 build a mathematical model that is so perfectly 00:19:17.119 --> 00:19:20.720 tuned and so flawlessly adjusted for every conceivable 00:19:20.720 --> 00:19:23.960 physical variable that it actually reaches 100 00:19:23.960 --> 00:19:27.660 %? That is the question. Or is there a fundamental, 00:19:27.960 --> 00:19:30.900 deeply unquantifiable human element to sports, 00:19:31.099 --> 00:19:34.220 a literal madness born of pressure, adrenaline 00:19:34.220 --> 00:19:37.240 and sheer random chance that pure mathematics 00:19:37.240 --> 00:19:40.400 will simply never be able to solve? That is exactly 00:19:40.400 --> 00:19:42.220 the kind of lingering thought we'd love to leave 00:19:42.220 --> 00:19:44.779 you with. A massive thank you to you for joining 00:19:44.779 --> 00:19:46.839 us on this deep dive today and exploring the 00:19:46.839 --> 00:19:48.660 numbers behind the madness. We hope you walk 00:19:48.660 --> 00:19:50.380 away looking at the statistics, the narratives, 00:19:50.460 --> 00:19:52.099 and the underlying systems in your own life a 00:19:52.099 --> 00:19:53.859 little bit differently. Until next time, keep 00:19:53.859 --> 00:19:56.319 questioning the accepted metrics, keep looking 00:19:56.319 --> 00:19:59.160 for your own 2 % edge, and we will catch you 00:19:59.160 --> 00:19:59.799 on the next one.