Brian Booker

Who's winning + where to sleep. Game predictions, hotel rate data & booking guides for fans chasing their team on the road.

[Brian Booker @thebrianbooker]

the market beat all seven of our models over the weekend. 51 graded games, favorites went 29-19-3 and our best three landed at 27-21-3

since 7/28 it's 363 graded events, m4 back in front at 215-137-11 (61.1%), favorites 210-142-11 (59.7%), that gap was 2.9 points on the 12th and it's 1.4 now

m3 and m6 are sitting on the exact record you'd get just picking every favorite, 210-142-11

[Brian Booker @thebrianbooker]

nfl preseason engine sitting at 10-6-1. 3-0 on the 14th, then 4-3 on saturday's seven

the 13th our loudest call lost, 49ers at 67.7%.. saturday the top two both came in though, ravens 71.5% won 24-7 and bills 69.6% won 29-14

three we got wrong were giants 62.8%, seahawks 61.8%, chiefs 53.4%

jaguars were the only road team we picked all night at 56.8%, won 24-20 in new orleans

[Brian Booker @thebrianbooker]

two ugly days on the board. 28 graded games across mlb and wnba on the 13th and 14th, and our seven models came in between 10-18 and 12-16

market got the same treatment, favorites went 12-16, dogs took 16 of the 28

since 7/28 it's m2 and m4 tied at 188-116-8 (61.8%), favorites at 181-123-8 (59.5%). gap was 2.9 points two days ago

[Brian Booker @thebrianbooker]

nfl engine's first real preseason night went 2-3-1, and the tie was an actual 13-13 final in colts-patriots

loudest call was the 49ers at 67.7% and they lost to the titans 19-13.. the two wins were basically coin flips, steelers at 50.8% and bengals at 54.6%, so that's 3-3-1 overall counting canton

it's preseason, mostly backups playing for jobs, not reading much into the misses or the hits yet

[Brian Booker @thebrianbooker]

nfl engine's first real preseason night went 2-3-1, and the tie was an actual 13-13 final in colts-patriots

loudest call was the 49ers at 67.7% and they lost to the titans 19-13.. the two wins were basically coin flips, steelers at 50.8% and bengals at 54.6%, so that's 3-3-1 overall counting canton

it's preseason, mostly backups playing for jobs, not reading much into the misses or the hits yet

[Brian Booker @thebrianbooker]

@ant0ni0_r0mer0 CLAUDE

[Brian Booker @thebrianbooker]

284 graded events since 7/28, 48 added over the last three days

m2 now leads at 177-99-8 (64.1%), m1 m4 and m5 all one game back at 176-100-8, favorites over the same stretch 169-107 (61.2%)

m4 led last update at 65.8% then went 26-22 over the new 48, worst of the seven, and lost the lead

where the models split from the market, m2 is 26-17 (60.5%), best of the table

[Brian Booker @thebrianbooker]

@jack_9947 skills

[Brian Booker @thebrianbooker]

models split with the market on three games tuesday and took two.. cardinals +150 (5 of the 7 liked it) and marlins +103, all seven on that one. the miss was red sox +120, also all seven there. board was 18 graded across mlb and wnba, best models went 10-8, favorites 9-9

[Brian Booker @thebrianbooker]

the market took monday. favorites went 8-4 on the august 10th board across mlb and wnba, our models went 6-6 to 7-5, and all three games where we split from the market went the market's way - cardinals +106 (all seven models on that one), orioles +101, giants +130

since july 28 m4 still leads, 156-84-8 (65.0%) vs 148-92 (61.7%) picking favorites, same 248 graded games

[Brian Booker @thebrianbooker]

@aftahi_ai @grok what is the statistical probability that this is true?

[Brian Booker @thebrianbooker]

Since 7/28, our best performing model predicting outcomes of 236 graded sporting events is:

the m4 model:
w/l/d
150-78-8 (65.8%)

m4 went
24-15 against the market (61.5%)

the favorites on the 236 games was 141-87 (61.8%)

all models pictured

[Brian Booker @thebrianbooker]

28/ data sources for all of this - nflverse, the open community nfl data project (rosters, snap counts, 17 years of injury reports, CC-BY license so shoutout to them), espn's public apis for schedules and results, and our own hand built coaching ledger verified against team announcements. everything here is free public data

[Brian Booker @thebrianbooker]

1/ we've been building an nfl prediction engine from scratch this week, moneyball style, the whole idea is that a team is just the sum of its players. this thread is every factor we tested and what actually survived, and TBH some of it surprised us

[Brian Booker @thebrianbooker]

27/ one live receipt already, the engine's first real call went out last night, panthers 53.4 in canton, and they won 33-30 on the last snap of the game.. one preseason game proves nothing tbh, but the record is 1-0