Read on Twitter
Some people think that sports betting can only be beaten by using computer ratings and models. On the flip side, there are others who think that computer models are completely useless.
Both of these groups are wrong.
Thread...
Both of these groups are wrong.
Thread...
This is something I've been thinking about for a while. I just never sat down and put it into words.
First of all, any computer model or system is only as good as the data being put into it, and the person (or people) who created it.
First of all, any computer model or system is only as good as the data being put into it, and the person (or people) who created it.
There's no computer rating system that's ever been able to consistently beat closing lines, at least in popular sports where there's a lot of liquidity and line movement.
If so it's being kept secret and will probably remain that way forever, but imo it's not even possible.
If so it's being kept secret and will probably remain that way forever, but imo it's not even possible.
It's been proven time and time again that closing lines are the best predictor of the outcome of events.
With this being the case, it would be logical to think that computer ratings are actually useless, but that's not the case.
With this being the case, it would be logical to think that computer ratings are actually useless, but that's not the case.
I've mentioned a few times that I use Jeff Sackman's Elo ratings for tennis.
While far superior to ATP and WTA rankings, the rating system still falls short when it comes to predicting winners, which Jeff has acknowledged before.
Why is this?
https://tennisabstract.com/reports/atp_elo_ratings.html
While far superior to ATP and WTA rankings, the rating system still falls short when it comes to predicting winners, which Jeff has acknowledged before.
Why is this?
https://tennisabstract.com/reports/atp_elo_ratings.html
Well, because there are numerous factors in tennis that can't be included, or properly accounted for, in the ratings.
Just off the top of my head...
Current form
Court speed
Weather
Fatigue
Accumulated fatigue
Head to head
Motivation
Injuries
Type of ball
And probably more.
Just off the top of my head...
Current form
Court speed
Weather
Fatigue
Accumulated fatigue
Head to head
Motivation
Injuries
Type of ball
And probably more.
Models are fantastic at telling you who were the better players on certain court surfaces over the last 52 weeks, or 5 years, or whatever.
But when it comes to predicting the winner of today's match between Serena Williams and Shelby Rogers... not so much.
But when it comes to predicting the winner of today's match between Serena Williams and Shelby Rogers... not so much.
So what exactly are ratings good for?
In short, saving time.
Remember when I said they are superior to ATP and WTA rankings? Well, right off the bat that gives you a better starting point to begin your analysis. Plus they can be sorted by court surface as well.
In short, saving time.
Remember when I said they are superior to ATP and WTA rankings? Well, right off the bat that gives you a better starting point to begin your analysis. Plus they can be sorted by court surface as well.
So when the Elo ratings tell you Player A should be -150 vs. Player B, yet the line is reversed, you can be pretty sure that Player B has an edge in one or more of the other factors I listed above.
From there it's up to you to determine if that difference is justified or not.
From there it's up to you to determine if that difference is justified or not.
This is how I handicap college football. I use a few rating systems to see how they compare to current lines, and from there I try to figure out why they don't line up.
If there's a big difference, it's almost always due to QB injuries, which aren't included in rating systems.
If there's a big difference, it's almost always due to QB injuries, which aren't included in rating systems.
When you understand what inputs are used in what models, over time you can almost predict what they will spit out.
There's one model I use for CFB that's superior to all others imo. I won't say it here, but you can DM me for details. I'm hoping it remains free for a while.
There's one model I use for CFB that's superior to all others imo. I won't say it here, but you can DM me for details. I'm hoping it remains free for a while.
I also use a combination of different rating systems for FCS college football as well.
@JeffDonchess probably didn't know that DRatings is one of the ones on my list, but he does now.
Hopefully we have a few FCS games this fall because that level is more beatable than FBS.
@JeffDonchess probably didn't know that DRatings is one of the ones on my list, but he does now.
Hopefully we have a few FCS games this fall because that level is more beatable than FBS.
Hopefully this gives you an idea of what's actually going on when people say, "My model has Team A -250" when the line is -150.
Well, does this model know that Trout is out today? What about Team X's tired bullpen? How about the wind blowing in, or other ballpark factors?
Well, does this model know that Trout is out today? What about Team X's tired bullpen? How about the wind blowing in, or other ballpark factors?
This goes back to what I said in the beginning that computer ratings are only as good as the data and the creator(s).
If the model accounts for every factor, then that's great, but it's not really possible without a full staff monitoring every little detail around the clock.
If the model accounts for every factor, then that's great, but it's not really possible without a full staff monitoring every little detail around the clock.
