Use of GAT For Pick 3

G.A.T. Engine general discussion
mitachoo
Casual
Casual
Posts: 12
Joined: Thu Feb 18, 2016 4:57 pm

Re: Use of GAT For Pick 3

Post by mitachoo » Wed Jun 21, 2017 12:34 am

To answer your questions from my previous post:

The hits column is how many in-line matches I got.

Yes, pool is the number of combinations played.

Only a straight win can return an outcome.

mitachoo
Casual
Casual
Posts: 12
Joined: Thu Feb 18, 2016 4:57 pm

Re: Use of GAT For Pick 3

Post by mitachoo » Fri Jun 23, 2017 1:24 am

Here's the backtesting results, for Pick 3 (Day) and Pick 3 (Night) Connecticut.

https://sites.google.com/view/tamp/home

The problem I have is this: although on the face of it the results seem WAY better than Random Play, how the hell do I know that my backtesting window didn't just happen to start in the middle of a hot streak? I've been burned by this many times before . . . I found what I thought was a great strategy, only to find it would peeter out and even out eventually.

Note that "Hit Rates" are Vertical only (i.e. within position).

mitachoo
Casual
Casual
Posts: 12
Joined: Thu Feb 18, 2016 4:57 pm

Re: Use of GAT For Pick 3

Post by mitachoo » Fri Jun 23, 2017 6:06 pm

I looked at the numbers in more detail and there are some interesting things here. Again, this assumes this is NOT a “temporary” hot streak that will eventually die; if it is, all bets are off.

Looking at the Binomial Distribution Function, the “Pool of 3” stands out for both the Day and Night drawings. For Day, BDF says there’s only a 1.4% chance one would get 8 successes in 13 trials of an event of probability of 3/10. The number for Night is higher (6.5%) but still awfully low. My take-away: for some reason, the combination of GAT and how I’m processing the data seems to “favor” a “Pool of 3”.

I looked at all the Jackpot wins between Day and Night and found that most contain at least 1 “Pool of 3”’. Presumably one could use this as a “trigger” and only play drawings that have at least one “Pool of 3”. Here are the before and after financials if one were to do have done this with the back-tested data. The net profit goes down from 2729 to 2232, but the ROI increases from 215% to 291%.

BEFORE:
Day: spent 715, won 2500, net 1785 (ROI 250%)
Night: spent 556, won 1500, net 944 (ROI 170%)
Total: spent 1271, won 4000, net 2729 (ROI 215%)

AFTER:
Day: spent 324, won 1500, net 1176 (ROI 363%)
Night: spent 444, won 1500, net 1056 (ROI 238%)
Total: spent 768, won 3000, net 2232 (ROI 291%)

I should mention here that the “Pool of 1” for the Day stands out as well, at 3 success in 7 trials, BDF 2.3%, but in this case it is NOT consistent with Night (Night is at 24.3%). It’s of no use as a trigger anyway as only 1 of the Jackpot wins contains a “Pool of 1”.

Unless someone can convince me otherwise, I think I have to go further back in time back-testing wise to see if these numbers hold-up longer term, to confirm my window is not smack-dab in the middle of a “Hot Streak”.

Post Reply

Who is online

Users browsing this forum: No registered users and 17 guests