Pattern Finder

[ruthlessimon]

So far it has found 53 patterns, ranging from 8% to 71.2%.

The problem is, statistically, it's guaranteed to chuck out what will look like amazing patterns.

If I hire enough coin flippers, someone will hit 100 heads in a row

Worth keeping in mind - & for god sake don't eat cheese

[BetScalper]

So far it has found 53 patterns, ranging from 8% to 71.2%.

The problem is, statistically, it's guaranteed to chuck out what will look like amazing patterns.

If I hire enough coin flippers, someone will hit 100 heads in a row

Worth keeping in mind - & for god sake don't eat cheese

[Euler]

One of my favourite sites: -

http://www.tylervigen.com/spurious-correlations

[BetScalper]

It finally finished data crunching.

As most suspected it spat out allot of rubbish.

But I think it has identified 4 meaningful, unrelated patterns which when combined would give a 64% strike rate overall with average odds of 2.86 in-running.

So, in the morning I will create the patterns using 4 triggers to be run on each market using minimum stakes and see what occurs.

There are 2 BTL triggers and 2 LTB triggers. They all run at the same time.

Parameters are:

- Race Type
- Number of Runners
- Book% (Back and Lay)
- Price gap between 1st, 2nd, 3rd and 4th Favourites
- Race Distance
- Betfair SP
- WOM using Inner Price weighting only
- Selections Volume % over x period (Increase/Decrease)

Obviously it could all end in tears and the patterns not appear again until next year but it's been a worthwhile exercise, I think!!

[Euler]

Use data but model a market to gain the best edge. You are trying to anticipate something that is about to happen. So if you model why something happens you will see the set up occurring and getting more likely. At that point, you jump in slightly ahead of everybody else.

[Korattt]

One of my favourite sites: -

http://www.tylervigen.com/spurious-correlations

now that's kinda cool

[marksmeets302]

So, in the morning I will create the patterns using 4 triggers to be run on each market using minimum stakes and see what occurs.

Hopefully you've set aside some test data, not used in finding the patterns. Then you can now test the patterns on unseen data. Just out of curiosity, how often do these patterns show up? That number alone should give you some idea of the validity.