Arbusers

Does anyone here have good experience with rating systems?
Masseyratings seem to be very popular for a long time since it has been around for long. When i search information about how accurate it is, i can't find any hard data. Is there a good source to find historical bookmaker odds, otherwise i could maybe make some AI try to look it up for past games and put together some track record.

Barttorvik is also good i heard but for college basketball.

Then other systems like dratings i looked into but the accuracy seems worse then the actual bookmaker odds.

And if none of these systems is better then bookmaker odds, then what is the point of it all?

A little bit off-topic, if I may.
I'm a big fan of building Stats/Math-based original metrics/ELO/whole systems/formulas of any kind, I really like the original thinking and ideas created by Minds like Kenneth Massey & Ken Pomeroy, I salute those creative Minds, but at the same time, I think it's much better for everyone to consider and create their own unique ones of any kind and develop them constantly in the timeline. Much better.
Also, there is no existing AI software better than someone's creative Mind, believe me.
Last but not least, the pure Enjoyment of your own creative process and testing and improving of your own invented instrumentations is much better than using someone else's product/software/formulas/ELO/systems, created by other people.
Just my 2 ¢.

arb12 wrote: ↑

Sun Apr 13, 2025 1:08 am

A little bit off-topic, if I may.
I'm a big fan of building Stats/Math-based original metrics/ELO/whole systems/formulas of any kind, I really like the original thinking and ideas created by Minds like Kenneth Massey & Ken Pomeroy, I salute those creative Minds, but at the same time, I think it's much better for everyone to consider and create their own unique ones of any kind and develop them constantly in the timeline. Much better.
Also, there is no existing AI software better than someone's creative Mind, believe me.
Last but not least, the pure Enjoyment of your own creative process and testing and improving of your own invented instrumentations is much better than using someone else's product/software/formulas/ELO/systems, created by other people.
Just my 2 ¢.

I don't disagree. But i think my problem is: i'm really sharp when it comes to prediction markets like economics, politics, culture, etc. Not when it comes to sports, in fact i like never even watch sports. (ball sports, which most sports are) But now i'm getting into MMA and Boxing as a niche specialization. F1 is also a sport i know something about.
There are some people i follow who are really good at it so i'm trying to learn to do it myself, the handicapping. (I'm talking about MMA) I do realize much also is like a decade long experience, you need to know all the little details that can influence the outcome. So it's kind of easy to say build a model but i think to get it on point you can easy spend 5-10 years or maybe fail and it never beats closing line values.
Usually i bet on quite niche things, like polymarket has many special events up. Eurovision is something i'm now trading a lot, mostly betfair but liquidity is shit for now.

Anyway to get on topic, do you know more about how these models are performing?

thepredictiontracker.com/

This site seems to compare different prediction systems for different sports and years. You can see how it performed. Seems like Masseyratings is underperforming against the bookmaker spread. But on the other hand i also noticed for like the majority of games it's pretty much the same probabilities as the bookmaker odds, or very close. So then it's also not that weird to lose money because you lose the spread each time betting.

Also how do they measure this, against what bookmakers and such.