How to simulate games with actual conditions?

**hutennis** · 01-05-13, 01:20 PM

Originally posted by livinglegend

I am interested in a theory that if you do an accurate simulation enough times, you will have an excellent chance to win, so you can have long term +EG. Does anyone know of a program/website that allows this? Whatif sports is the same idea, but unfortunately they don't have current rosters. Thanks!

It looks to me like you are missing a pretty important point.

Your simulation/analysis/model does not exists/performs in a vacuum.
It can not be accurate just by itself.
It is either MORE accurate or LESS accurate than something it competes against.

For you to make money, your simulation/analysis/model has to be better than whatever simulation/analysis/model is being used by whomever takes an opposite side of your bet.

If you would bet against your next door neighbor whose model is to make money line picks by flipping coin, then it is not too hard to come up with model that would be better than his model and make a killing in a process.
I don't think, however, you'd be betting against this type of models very often.

**livinglegend** · 01-05-13, 02:06 PM

That's definitely a valid point, but I was thinking more in terms of futures, moneylines, and even props in some cases, where if you can pick the correct winner a high % of the time, you can make a fair amount of money. I was just wondering if these programs exist, and if they do, whether they exist in an up-to-date form.

**hutennis** · 01-05-13, 03:06 PM

Originally posted by livinglegend

That's definitely a valid point, but I was thinking more in terms of futures, moneylines, and even props in some cases, where if you can pick the correct winner a high % of the time, you can make a fair amount of money. I was just wondering if these programs exist, and if they do, whether they exist in an up-to-date form.

It does not mater whether you think in terms of futures, MLs props or anything else.
Picking the correct winner a high % of the time is not enough to make money.
To make money you must pick a winner with %% that is higher than break even %% suggested by the odds you take on your bet.

How do you plan to get to that point?
Why would anyone with a method capable of producing such a results would make it available to the public
in a form of software, simulator etc. ?

**HUY** · 01-05-13, 05:52 PM

Originally posted by livinglegend

I am interested in a theory that if you do an accurate simulation enough times, you will have an excellent chance to win, so you can have long term +EG. Does anyone know of a program/website that allows this? Whatif sports is the same idea, but unfortunately they don't have current rosters. Thanks!

This is not a theory, this is plain nonsense.

**Miz** · 01-06-13, 12:51 AM

The bottom line is... you have to accurately calculate the probability of something happening, and you have to do it significantly better than whoever you are betting against. To do that takes:

1. Ideas that intuitively make sense as the basis of your model or sim (e.g. rushing yards per attempt can help predict the winning margin of a CFB game)
2. Your output has to be tested in an unbiased fashion and without the test and training set being the same
3. You have to test enough to generate a statistically significant sample of output data
4. Your test results have to show that you can predict things well enough to bet at a profit large enough to justify your time doing it

Best of luck. We all start somewhere. Steps 1-4 take hard work to achieve.

Nobody is going to be able to point you to a website sim that can provide this valuable information. People have tried to use Sagarin ratings, Dunkel, etc. They may work for a short time, but they eventually get priced into the market. So if someone created a golden goose (i.e. successful model), they aren't going to share it with you.