Closed calexil closed 7 years ago
That is from my fork and its a console printout. The Weights are different. The issue history alludes i been messing with the chromosomes.
crap, should I not be using it? I merged it into my master lol
I have backups.
a simple yes use it or no dont would be appreciated.
I am not a smart man
the current setup uses the master, your chromo, and the cleaned 86k records
if thats wrong Ill revert to my old chromo
Don't use it. Its a develop branch.
okie day, ill revert it(the chromo)
If you want to know what chaos using that chromosome with its new time weights bring. Assuming your using the main branch code, the weights are effectively +1 size and everything is offset where the changed line begins. lol
oh dang, okay I reverted to my old chromo for the master and my pr
still doesn't answer my original question tho ^ ^ , but if people dont wanna give away their strategies I will understand
Since I'm working with a clean(er) record set would it make sense to just reset my pool, or let it go as is?
Use multiplier depending how much you want to trust the chromosome when it guesses wrong. Trust is up to you.
Personally i keep it 0.4 or lower when i want both slow burn gains AND not flood odds with my ever increasing total(although there will be a point it will be forced to bet magnitudes higher then everyone else with the current scale).
So do this: Restart pool (purges all chromosomes), (optionally import a working chromosome to feed in the new pool), run sim until you're happy with its results or it stopped changing significantly.
Actually on the current system, constantly running the same chromosome a long time, as 100ths of new matches are added could be bad. Hard to explain, it involves how its gets locked into its old solution space that new constantly added game data cannot be trained in as it cannot escape its weights.
so there really is no good slow burn afk betting solution?
You understand how this bot works from the documentation provided in the GUI?
If understanding how betting, probability, odds, and time work. If the chromosome is correct. If the match data is correct. If understanding how scientist mode works. If understanding how its confidence value changes how much it bets. If understanding how it selects a character by scoring between then using the chromosome weights and their data. If you understand the multiplier applies just before the bet button is hit. Then if you made a graph of your gains over time, it will be steady.
Think of betting as exchanging 10% of what you have into odds as a returned %, if correct. If you been using 2.2 multiplier, think how it affects betting your total? What are you losing at a poor guess? 5-20% of everything? How much time did it take your bot to get that much lost? Low burns is using very low amount of your total as you bet. I don't even know what the bot defaults as % using 1 scaling.
You might ask is the chromosome correct? Its testing against and predicting everything in the records. The % should say how much was predicted correctly, the money should say how playing with confidence and betting upsets correctly against historical odds earned even more. (although here is what i do not fully understand how it does upsets).
We must Summon @synkarius !
I think I should just slowly lower my multiplier. But I honestly dont know.
I was at neutral gains all night last night, today I dropped it from 2.2 to 1.9 and Immediatly started climbing again
need to find the sweet spot
I thing lowering my multiplier every 5 mil by .2 will be my new strat
I am totally mystified, even after reading the guide and using this thing for many months.
I currently bet on everything, at a 2.2 multiplier, and I am using the newest cleanest data(despite 0 times constantly being added and muddying it)
my current chromosome chromosome--2017-3-9-21.3.txt
initially: every month or so I update the genetic weights(which currently seems broken btw)
Should my multiplier be higher? Should I enable aggressive betting?