Best Stats for Predicting CFB Games

[Tanko]

I'd like to leverage some of the SBR poster's wisdom on CFB.

I have "scraped" a lot of CFB stats off the internet into Excel for the 2014, 2015, & YTD 2016 CFB seasons and used some of them (e.g. YPP comparisons) to generate predictions but have not really done much better than 50/50 on games. I do not have enough time to run a regression or back-test the potentially hundreds of combinations of stats to find the right system.

Can anyone provide insight into which factors you believe are most critical and how they should be weighted? Here are the stats I have for every game over the period I mentioned. As you can see the stats are shown for each team for ALL GAMES and for RD/HOME games (depending upon if they are on the road or at home for the given week). There is a legend explaining the acronyms at the bottom.

I'd be more than happy to take your suggestions and provide results of some back-testing as time allows. If you have a system to suggest, please be relatively specific/detailed in how you think it should be used to predict scores that can be compared to the spreads.

I would have attached this in Excel format but there seems to be a problem with uploading files on SBR (error message: Exceeds 2.9GB forum limit).

	Off Avg Score	(all games)
	Off Avg Scr	Rd/Hm
	Opp Def Scr Allow	Avg
Defense Overall	Def Scr Allowed	(all games)
	Def Scr Allowed	Rd/Hm
	Opp Off Scr Made	Avg
Turnovers	TO	(all games)
	TO	Rd/Hm
Offense First Downs	Off 1st Downs	(all games)
	Off 1st Downs	Rd/Hm
	Opp Def 1st Dn	Avg
Defense First Downs	Def 1st Dns Allow	(all games)
	Def 1st Dns Allow	Rd/Hm
	Opp Off 1st Dn	Avg
Time of Poss	TOP	(all games)
	TOP	Rd/Hm
Offense Overall Stats	Tm Off Yds	(all gms)
	Tm Off Tot Yds	Rd/Hm
Offense Rushing Stats	Off Rush Yd/gm	(all games)
	Tm Off Rush Yds	Rd/Hm
	Opp Def Rush	Avg Yds
	Off Rush Attmpts	(all games)
	Off Rush Attmpts	Rd/Hm
Offense Passing Stats	Off Pass Yds	(all games)
	Tm Pass Yds	Rd/Hm
	Opp Def Pass	Avg Yds
	Off Passing Atmpts	(all games)
	Off Passing Atmpts	Rd/Hm
	Off Pass %	(all games)
	Off Pass %	Rd/Hm
Defensive Overall Stats	Tm Def Yds	(all gms)
	Tm Def Tot Yds	Road/Hm
Defensive Rushing Stats	Def Rush Yd/gm Allow	(all games)
	Tm Def Rush Yds	Rd/Hm
	Opp Off Rush	Avg Yds
	Def Allows Rush Attmpt	(all games)
	Def Allows Rush Attmpt	Rd/Hm
Defensive Passing Stats	Def Pass Yds Allow	(all games)
	Tm Def Pass Yds	Rd/Hm
	Opp Off Pass	Avg Yds
	Def Passing Atmpt Allowd	(all games)
	Def Passing Atmpt Allowd	Rd/Hm
	Def Pass % Allow	(all games)
	Def Pass % Allow	Rd/Hm
Offensive YPPT Stats	Off (all gms)	YPPT
	Off Rd/Hm	YPPT
	Oppon Def	Avg YPPT
Defensive YPPT Stats	Def (all gms)	YPPT
	Def Rd/Hm	YPPT
	Oppon Off	Avg YPPT
Offensive YPP Stats	Off (all gms)	Avg YPP
	Off Rd/Hm	YPP
	Oppon Def	Avg YPP
Defensive YPP Stats	Def (all gms)	Avg YPP
	Def Rd/Hm	YPP
	Oppon Off	Avg YPP
Offensive YPR Stats	Off (all gms)	YPR
	Off Rd/Hm	YPR
	Opp Def	AvgYPR
Defensive YPR Stats	Def (all gms)	YPR
	Def Rd/Hm	YPR
	Opp Off	Avg YPR
Offensive PYA Stats	Off Pass (all gms)	PYA
	Off Rd/Hm	PYA
	Opp Def	Avg PYA
Defensive PYA Stats	Def (all gms)	PYA
	Def Rd/Hm	PYA
	Opp Off	Avg PYA

Yards - Total yards per game gained or allowed.

YPP - Average of total yards gained (or allowed) per play.

TO - Number of turnovers committed (for offensive stats) or allowed (for defensive stats).

FD - First downs gained or allowed per game.

TOP - Time of possession for either team or their opponent.

ATT-YDS (under rushing) - Per game averages - rushing attempts, rushing yards gained (or allowed).

ATT-YDS (under total) - Per game averages - total plays, total yards gained (or allowed).

YPR - Rushing yards gained (or allowed) per rushing attempt.

CMP-ATT - Passing completions and attempts per game. (Either by team or by opponent.)

PCT - Team's passing completion percentage or completion percentage surrendered by their defense.

YDS (under passing) - Per game averages of passing yards gained (or allowed).

YPA - Passing yards gained (or allowed) per passing attempt.

YPPT - Average of total yards gained (or allowed) divided by the points scored per game (or allowed).

INT - Interceptions thrown (offensive stats) or team interceptions (defensive stats).

FL - Fumbles lost (offensive stats) or fumbles lost by opponents (defensive stats).

TO - Total turnovers lost (offensive stats) or turnovers lost by opponents (defensive stats).

DIFF - Team turnover differential represented as a per game average. + means more turnovers forced than lost.

#-MADE - Per game averages of 3rd (4th) down opportunities and 3rd (4th) downs converted.

PCT (Under 3rd, 4th down conversions) - Interceptions thrown (offensive stats) or team interceptions (defensive stats).

KR-YDS - Kickoff returns and kickoff return yards as per game averages. (gained or allowed)

YPKR - Kickoff return yards per return. (gained or allowed)

PR-YDS - Punt returns and punt return yards as per game averages. (gained or allowed)

YPPR - Punt return yards per return. (gained or allowed)

PEN-YDS - Team (offense) and opponent (defense) penalties and penalty yards per game.

[HeeluvaGuy]

I'm not sure I can help with your request, but I would love to see your penalty data. I haven't found a good site that has penalties by offense vs. defense.

[Tanko]

I'm not sure I can help with your request, but I would love to see your penalty data. I haven't found a good site that has penalties by offense vs. defense.

Apologies for the error HG, but although PEN-YDS is shown in the legend table, I do not have the data in the stats table.

[sportsfan9698]

None of the stats are of any value at all. Between the linesmaker and 1000's of wagers and dollars wagered all of that is baked in. total waste of time and effort, which will usually point you in the wrong direction.

1) One way or another determine which team has the greater motivation to be prepared to win.
2) Double check #1 to make sure you have it right - then wager with confidence.

[Tanko]

None of the stats are of any value at all. Between the linesmaker and 1000's of wagers and dollars wagered all of that is baked in. total waste of time and effort, which will usually point you in the wrong direction.

1) One way or another determine which team has the greater motivation to be prepared to win.
2) Double check #1 to make sure you have it right - then wager with confidence.

I'm not trying to be offensive, but I'm curious. Based on your post, why do you visit SBR? Your comments,... "Total waste of time and effort usually point you in the wrong direction" seeem to indicate you believe sports betting is a crap shoot and nothing more. While I can't argue with your opinion, why would you look for insight on Sports Betting Preview?

Thanks for your reply.

[sportsfan9698]

I'm not trying to be offensive, but I'm curious. Based on your post, why do you visit SBR? Your comments,... "Total waste of time and effort usually point you in the wrong direction" seeem to indicate you believe sports betting is a crap shoot and nothing more. While I can't argue with your opinion, why would you look for insight on Sports Betting Preview?

Thanks for your reply.

I'm not trying to be a smartass but you clearly have not given this much thought. Just let it simmer you will start to understand my point of view, which is similar to many here. Some prefer the technical number crunching... it's all good

[sportsfan9698]

Think about stock traders.... some like fundamental analysis, some prefer technical analysis

[Tanko]

BTW SF9698, the idea of selecting the team with greater motivation to win is a great suggestion but is tough to measure. A lot of subjectivity. But, I like the 2nd part. Once you select a team, wager with confidence.

[sportsfan9698]

A perfect example... several years back Michigan State was preparing to play Michigan. MSU had already been dominating the series like 5 of the past 6 years. But Michigan program can get big heads (like all teams) and one of the defenders was laughing when being interviewed and called MSU "little brother". Dantonio made sure that shit was all over the locker room. Michigan got smashed when all of the stats said they should have won easily. Dantonio has been using that for years now... series is now like 10 of the last 11 MSU

This year will likely be different... but who knows ATS

[sportsfan9698]

Another great example... Mr Jefferson in Fast Times at Ridgemont High

[ScreaminPain]

One thing I see that might make a difference to your stats. SOS is a variable in CFB that may slant some of your figures. I see you mentioned YPP- I'm not certain if you mean Yds/pt or Yds/play, but either way SOS may alter your computational output.

Example: What if team A scores at the rate of 13 Yds/pt against team B that typically allow teams to score at 11 yds/pt..... and now are playing team C whos defense forces teams to score at a rate of 16 Yds/pt. What will the rate be for team A in this scenario?

Of course you'll need to determine the predicated Yardage of both teams before you could apply any SOS calculations....

[Tanko]

One thing I see that might make a difference to your stats. SOS is a variable in CFB that may slant some of your figures. I see you mentioned YPP- I'm not certain if you mean Yds/pt or Yds/play, but either way SOS may alter your computational output.

Example: What if team A scores at the rate of 13 Yds/pt against team B that typically allow teams to score at 11 yds/pt..... and now are playing team C whos defense forces teams to score at a rate of 16 Yds/pt. What will the rate be for team A in this scenario?

Of course you'll need to determine the predicated Yardage of both teams before you could apply any SOS calculations....

Thanks for replying.

Great point. SOS is another factor I use in predictions. I refer to it as Avg Opponent Power Ranking (AOPR). Although I did not show it on the table of stats, it is one of the pieces of data that I collect for each game. Any suggestions on how to weight that factor vs others?

The YPP is yards per play while YPPT is yards per point. I have both.

[gojetsgomoxies]

i've always wondered if you get two teams - none extreme stats - where there is a mismatch between either rushing offense and opponent D or same with passing if that is an opportunity. basically i would think it matters whether a 65th ranked team - arizona say - has a chance to stop for mccaffray for stanford. if they can then they have great chance in game, if not then gonna get killed.. and of course much of the game is passing these days so teams aren't generally as focussed on stopping the run.

another thing i'm toying with...................... are underdog teams with a big variation between offense and defense ranking good bets? they keep it close either with a stellar O or D. probably thinking more D but not sure. stellar O is probably ok, because if you score at least 20 points it's not that easy for opponent to cover 20+ point spread. plus if you are moving the ball well that eats up time and other team won't have that many minutes to score before "protect the lead" with running plays begins.

you seem to have an idea of net turnover margin and/or YPP reversing.....

sort of similar to point made above but highly ranked teams with mediocre offenses.. are they a good fade?

[ScreaminPain]

Any suggestions on how to weight that factor vs others?

I use "League Average" as a gauge verses each teams SOS. There are various ways to evaluate this, but one simple method might like this:

SOS: Team "A"/((Team "A" + Team "B")/2)...using Sagarin's NFL figures it would look like this: Jax 16.4, Chi 15.7
therefore 16.4/((16.4+15.7)/2).....=Jax 1.02, Chi .98.

This means that Jax has faced a schedule that is 2% harder than Chicago.


((Team "A" * team "B")/League Avg.)*SOS

This is somewhat simplified, however the basic context illustrates how it may be used.

[JTrain]

Making a computer prediction system from scratch with any kind of edge ATS is extremely difficult and I wouldn't recommend it. There are a handful of systems that are capable of hitting 53-58% ATS and even a handful that perform in that range multiple years in a row. See here. Those are worth looking at.