Hey Kerry, first off I wanna say I’m loving the content. Been making a lot of money betting some of the plays your model has predicted and I wanna say thanks. I’m a positive EV bettor and I use an algorithm to scan the market and find lines where sportsbooks are messing up (getting a play at +180 on fanduel, when every other book in the world has the same bet at +130). Always looking for new ways to beat the books, as I’m sure you are as well. as you know, sportsbooks are billion dollar companies and have models and algorithms of their own and have paid oddsmakers who’s full time job is to accurately set and price lines. So my question is, what makes your model more accurate than a billion dollar company’s whos full time job is setting lines? Now I know, especially during playoffs, that the “true line” can be moved based on the public opinion and supply and demand to balance money on both sides of the bet (ex: Tatum’s point line opens at 26.5 and come gametime has moved to 28.5 because casual bettors love rooting for star player overs so his line gets artificially driven up to balance out the money on both sides) and we can take advantage of that, but was wondering if there’s anything you do that the sportsbooks aren’t ? Thanks Kerry and keep the videos coming my man 💰
The model doesnt account for injuries does it? I feel Like it would be better than 18-20 if you didn’t include games where star players have been injured
good stuff, you can also do the following team_df = team_df.astype({'G': int, 'OPts': int, 'DPts': int, 'ORtg': float, 'DRtg': float, 'Pace': float}) intead of writing in line by line :-)
Hi Cobb! Wow! That certainly makes sense and is much cleaner. Thanks for sharing! Btw, I'm still planning on posting a video on an NBA ML model. It will probably be posted in a couple weeks. Cheers!
Hi Epic! I'm glad you liked the video. I will post this on Google docs and provide a link in the description. I'll let you know when it is available in the comments here. Cheers!
Hi Tom! I'm glad you like the video. I agree, there are some really intriguing matchups in the first round. I would rather see the top players in the playoffs. But with some of them out (Giannis, Butler, etc.) these series take on a whole new outlook. Can't wait to see these games. Cheers!
Hi Charles! You're very welcome! I'm glad you're getting some Python tips from these videos. I'm learning new ways to do things in Python all the time. Cheers!
@@Cobbtrades I'm working on a ML model right now. I've created several machine learning sports models, I just haven't published any yet. I'll let you know when I post the video. Stay tuned!
Hey Kerry, just found your channel and I'm loving watching all of your videos. I've been learning Python as a hobby for a couple years but have a love for learning about sports data - so this channel is a god send for me currently. Not sure how much you're interested in the NBA but it would be cool to see something like scraping some NBA player data and then maybe querying to see who would be the top candidates for the various individual awards like MIP, 6MOY based on certain criteria etc - see if there are any outliers etc
Hey Tom! I'm glad you enjoy the videos. We are in the same boat. I love data and sports! Ironically, I am preparing to record a video on scraping NBA data today! So your request is very timely. With regards to player data, I have a scraper for that too. I'll need to run the code to make sure it still functions correctly. I recently read an article that claimed that the PIE statistic is very predictive of the season MVP award. We'll have to take a look at that. I'll come back here and let you know when today's video is available. Stay tuned!
Hi Tom! I just posted a video titled "Scraping NBA Data... With Python!". I'll be posting another video soon on stats that can be used to help predict NBA game outcomes. Cheers!
Happy Eclipse/Championship day everyone! I can't believe we are having a total solar eclipse *and* the NCAA Championship game ON. THE. SAME. DAY! Cheers!
Hello Kerry, I've enjoyed your recent videos on the KenPom Prediction Model. I'm seeing different results...on AL vs Connecticut, updated KenPom: Alabama 125.8, 102.6, 14.21, 72.8 PointDiff =1.9981 Win 57.2%, you show 23.1% Connecticut 126.7, 91.5, 11.31, 65 PointDiff = -1.9981 Win 42.8%, you show 76.9% Has the formula changed? I'm using the same one from your video on N.C. State vs UNC. Thanks!
Hi Cory! I'm glad you've enjoyed the videos. With regards to your calculations, it looks like the third numbers are off. The third number (AdjEM) is just the first number (AdjO) minus the second number (AdjD). In other words, AdjEM = AdjO - AdjD. So those numbers should be 23.27 for Alabama and 35.21 for Connecticut. While I did modify the formulas slightly, the win probabilities should be within 1% of numbers in this video. Let me know if this works out. Cheers!
Sorry to see a good thing coming to a close . However there's an 11 year old girl inn Raleigh who has beaten you chat gpt musks Grok and 25 or 30 million others at 109 and at 99.9% tile. That's what 10, ooo years of eyeballing for sorting evolution will do . Still enjoyed your process.
Hi Terence! It's been fun! I appreciate your comments and insight. I have no doubt an 11 year old can beat the entire field. I already have some ideas for next year. Looking forward to it. Cheers!
@@KerrySportsAnalyst it will be hard for us hunters to catch up with the gatherers . As a former teacher and mailman I can tell you in general they're better at the sort.
5 of the 8 have oers over 120 . State Clemson tenn. Do not . But tenn der has steadily improved. Uconn showing an oer of 127 . Is a new high . Offenses are soaring. Some defenses have not improved . You have some good brackets beating ai selections.
Hi Terence! Those are great observations. You're convincing me that great offenses are taking charge in the tournament. Tennessee has the best defense still alive in the tournament. We'll see how well they do against Purdue's #3 offense. Can't wait to see these games. I truly appreciate your support and comments! Cheers!
@@KerrySportsAnalyst my thinking now is the original historical minimums you originally highlighted are higher for offense now and defense minimums are higher . Bama for example Illinois another bother over 100 .
@@KerrySportsAnalyst my theory on states recent success is a little different having little to do with direct data . And more to do with uniqueness of lefthanded chirality or see cube roots . Dj burnes lefty center provides a weird inside spin and passing ability few teams can cope with . Even his own mates have had to adjust but think of a conch shell twist doing opposite of normal path integral. Tenn. A few years ago experimented with 9 lefties OK success but not major. Only like 12 % of pop is lefty . Now there are some drawbacks such as higher frequency of accidents and injury but for example who has most nba rings . 2 lefties Russell and Jackson.
Hi J9160! The probabilities in this video are for just winning the games. Check out my new video titled "A KenPom Betting Model" to see how to get spreads and totals from the KenPom stats. Cheers! ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-yHeWYDLmGos.html
Hi Aragorn! You are correct when it comes to the picks in this video (for the Sweet 16). However, in the first two rounds, KenPom predicted 5 lower seeds to win. It got 3 out of those 5 picks correct. I think this also points to how closely the selection committee aligns with the KenPom stats. Where the KenPom stats were different, KenPom got 3 of 5 correct. Cheers!
Hi Sack! Check out my new video titled "A KenPom Betting Model". It shows how to get the spreads, the totals, and the team totals using KenPom stats. Cheers!
Good job stats updated . The eyeball test can detect occasionally a chalk mismatch or a hotness factor. ( though I know stats people deny the later exist ) the eyeball may also be reflecting a jaundiced recency bias so check your self . 😊 Clemson shouldn't have beaten baylor by chalk but they did -defense and late game composure prevailed.
Hey Terence! I absolutely agree with you. While I present the data and the stats in my videos, I also look at experience, head coach, and recent wins against other tournament teams before making final decisions. Experience is huge! Cheers!
@@bigmombo3989 sorry I didn't see it but tigers from portal played more composed than bears from portal . Wildcats will be tested for a half but data says its a 10 plus win.
Hi User! I totally agree. Chalk makes KenPom look good. But KenPom did get 2 of 3 upsets correct though. Cheers! ❌(11) New Mexico over (6) Clemson ✅(10) Colorado over (7) Florida ✅(5) Gonzaga over (4) Kansas
@@ryanhemesath6686 Hi Ryan! Good point. When it comes to predictive models, when the stats used are much better for one team (Auburn) versus another team (Yale), the model will definitely pick the better team. But in a single elimination tournament, wild upsets will happen, for sure. Cheers!
I think Kerry's error here is that offenses have gotten stronger correspondingly defenses weaker . I think D weakness stretches the contender pool to those who have 98 or under and offense of 118 or better. With a single outlier . ( ? ) otherwise Kerry has really done the stat work. 4 of 6 teams he listed as contender don't fit even his parameters . Normalization - requires some adjustments that even Ken pom doesn't use .
Hi Terence! Thanks for your comment. It is very insightful. It would take a season or two for the stats to "catch up" to what is happening. With regards to 4 of 6 teams not fitting my parameters, I'm pretty sure they did when I made this video. Since KenPom stats are updated after each day, the stats might be different now. Cheers!
@@KerrySportsAnalyst Kerry don't take offense I think what you did is FABULOUS !!! HOWEVER I DID CHECK ON THIS YEARS # OF TEAMS SCORING 80 + POINTS VS LAST YEAR IT WAS LIKE 46 TO 23. REASONS BEING 1. Portal proven scorers . 2. More mature decision makers 3. Greater emphasis on 3 ball. As part of my "renormalization " of the data set YOU selected I just temporized to first 118 adjo then 120 adjo and upped the adjd to 98 . So my bracket picks are like 40 for 52 and yesterday's 8/8 . I think we are beating chat gpt and musks Grok and vegas models . So keep it up. My problem with the historical data set is this. Each game is an anomaly and each years data reflective of that year's set of anomalies. Yet there are strong threads of consistent outliers such as aztecs D . Does defense win championships ? And were a houston to win would it be more so for its adjd ? 🤔
Hi ZBatt! Check out my new video. Here are the KenPom records for the first four days of the tournament. Day 1: 12 - 4 Day 2: 9 - 7 Day 3: 7 - 1 Day 4: 6 - 2 That is an overall 70.8% win rate. Cheers!
Hey thanks for making an update video there is a lot of good data in here however I do have a suggestion. While I personally loved seeing the graph of upset seeds vs the high seeds and don’t necessarily want to see that disappear I’m wondering if you would consider showing historical data for the current matchups. For example you talked about 14 vs 6 having no wins and that being a bad sign for Oakland but it is actually 14 vs 11. I would love to see a chart for the historical results of those matchups as well. Again, thanks for putting out another video for this tournament. Looking forward to future updates.
Hey Gage! Thanks for the advice. I think that's a great idea and I will definitely put that in my next video. Oops, I misspoke about Oakland. My slides are correct though. Cheers!
With this iteration of the spreadsheet there is still potential for randomization to happen but as with any multi draw scenario the more you draw the more likely the higher probabilities are to be selected.
Hi Zachary! I totally understand the confusion here. These are conflicting stats. When this happens, I resort to the matchup. According to the KenPom prediction formula, Kentucky has an 86.2% chance of defeating Oakland. So I am picking Kentucky. Hope this helps!
Hi Priust! It's hard to not go with the heavy favorite to win it all. However, I am picking multiple champions since I'm entering 25 brackets in the ESPN Challenge. Cheers!
Hi MaverickFan! While I do not necessarily recommend paying for data, the annual subscription fee for KenPom is aroud $22. Also, you can see the pre-tourney data for 2024 before the games start today (3/21/2024). Cheers!
no computer needed - Samford sends kansas... home....time to comb your hairpiece mcneese over Zags... james madison over wisconsin... Grand Canyon sends St Marys home...
Hi Chris! I literally laughed out loud when I read your comment about combing my hairpiece. It might sound odd, but I appreciate that kind of humor. But as David Letterman once said, "if this was a hairpiece, why would I pick this one?" Lol. Cheers!
This is exactly what I was looking for, thanks for the analysis. It would be really interesting to do a similar analysis for upsets looking for statistical significance of the kenpom "four factors".
Hi King! I'm glad you like the video. I agree. I just posted a video on how to find upsets. It uses randomization of KenPom probabilities to pick a few upsets. We'll see how it works. Cheers!
Does the Kenpom formula adjust odds for style intricacies? As a for instance, take the Bama vs St Mary’s potential matchup. They have totally different tempos and play styles. Does Kenpom take that into account for the formula? Or is it just taking point spreads implied by adj em and basing probabilities on that?
Hi John! Yes, the adjusted tempo (AdjT) is included in the prediction formula. In fact, it is essential to determine how many possessions each team will have during a game. It then uses AdjO to predict the number of points each team will score. I'll try to discuss this in my next video. Cheers!
Something isn’t right. Using the methodology you outlined in the previous video(and at the top of this video to show the contenders), Tennessee should beat Purdue. What am I missing?
Hi Ryan! According to the video "Picking the Overall Champion" Purdue does not belong to the list of contenders. However, in a head-to-head matchup, Purdue is favored to beat Tennessee 57% to 43%. These are not necessarily conflicting since Purdue could beat Tennessee and then lose in the Final Four. However, since I am participating in the ESPN Challenge, I'm going to pick Purdue to be a champion in 1 or 2 brackets and Tennessee to be champion in 1 bracket. Trying to spread out my picks. Cheers!
