Data Manipulation
-
Hi Guys
Keen to hear as much or as little as anyone is willing to share about the process they go through for developing their own filters from their raw data. I have 2 filters cooking that are in their second phase. Theoretically Phase 2 is more improved that Phase 1, and the flimsy way I did this was to sort them each and every game in to those that would have been “successful trades” and those that would have been “unsuccessful trades”. I then took each column of data, and took an average of the percentages of each type of trade to find the average difference to see if I needed to exclude games more strictly - I hope this makes sense? If not, I will try better to explain it!
On Phase 2 this isn’t working, and the difference in average percentage is either non-existent has a negative difference - implying I maybe need to limit the top end of results too?....
Anyway, I don’t think this is effective and there must be better ways to improve my filters, perhaps different collections of data, perhaps excluding certain leagues...
What process do you all go through? Thanks for any input!
-
@darri
Hi Darri! Thanks for the reply!
What I will do is I will post more detailed info about the two filters to my trading thread on here (may not be today, probs at weekend) so you’ve got more info to work with! I appreciate any and all input!
-
Hi mate, its quite a big topic to summarise but if you can give me what your strategy is it would be easier to understand exactly what your after.
For me the process for filters/strategies is:
- Market overview, making the filter small in criteria to get a overview of how the market works. From here we can look at a number of things. Lets take my FHG project for example. All we tracked was : was there a FHG, what time was it scored, and on our stats filter what stats worked better, so in our case we inputted how many FHGs there had been in the last 10 games for the home and away side combined.
- Next you improve, from what im reading this is the stage you are at? So now with the data from before you cut out the nonsense and double down on what is working. So keeping on with the FHG, we saw that on average the home team was scoring more of the FHGs. So to improve on that point we now include a home scored in H1 into our filter. We then track this filter update for same amount of games as your last batch and see if it has improved, if it hasnt then the changes you made werent needed and you can try another one or accept that this is a solid strat.
- The next batch is the money batch, this is were you track price etc and build around your well tested filter.
Please let me know the strat you have and maybe even post the results sheet if you want. I also dont think you should overcomplicate this part. It really should be kept as simple as possible. Some people try to collect too much info that just isnt needed. What i do is i test but only try to tweak one area of the filter/strategy. This takes longer but it means your optimising one thing at a time and makes it better setup long term too. So take a second half goal example, if your filter had previously saw home team score more often, then on your next batch tweak on the home team metric to get that as optimised as possible forgetting about the rest, work on one area at a time when testing, do not rush it. You see people making 2-3 changes at a time an mis managing it completely. Ruining what could have been something solid.
Martin is right, leagues matter, for me as you know i only care about the top leagues, this is from experience but just avoid learning to test/trade in low leagues yes you will still have success but long term it will be void as youll never be able to trade them with bigger banks.
Leagues, The stats on your filter, goal times, home scored, away scored, ht score, ft score, odds, strike rate there is too much to narrow down and thats why we kinda need to know your strategy and then we can all help you to setup a sheet for better data collecting
-
@martin-futter said in Data Manipulation:
a couple of things off the top of my head
- Leagues matter, so yes I do look at which ones perform best after getting a lot of data.
- Does the filter work well in a different way? So maybe it was created to find a second half goal anytime but you have recorded goal times and noticed that there are lots of late goals so that would be the better angle. I think recording scorelines and goal times means you can go back later and test lots of different ideas.
Cheers Martin! Great input
-
a couple of things off the top of my head
- Leagues matter, so yes I do look at which ones perform best after getting a lot of data.
- Does the filter work well in a different way? So maybe it was created to find a second half goal anytime but you have recorded goal times and noticed that there are lots of late goals so that would be the better angle. I think recording scorelines and goal times means you can go back later and test lots of different ideas.
