Attached is a graphic from a particular NHL system that I am looking at right now. Starting with a zero balance, each point represents the balance at the end of each day regardless of how many bets were made. The results are +239 units on 1,064 bets over the last 5 NHL seasons and up until last week of this one. odds used are from Pinnacle. This was not found with random data mining, so I do not expect it to be an artifact of overfitting.
I cannot figure out whether this is a system that would be worth betting. While clearly outstanding in 2006 and 2007, it was unprofitable in 2008, and only slightly profitable in 2009 and 2010, and only very slightly profitable so far this year. If I plotted a curve of its results per year, it is on the downward slope.
It also requires quite a few bets to be made, sometimes >10 bets a day at certain points in a season, so bankroll issues would abound (though it doesn't appear to have particularly wild swings on a per-day basis.
Looking for thoughts on how to evaluate a system like this. Do you care more about the betting over the last two years, or does the 2006 and 2007 season seem like it might repeat at one point? Would you look for subsets of this data that show a higher return per unit bet, even if it means making less total money in the long run? I think everyone's looked at a chart like this at some point and wondered the same thing.
p.s. not a troll, not a tout, not selling this, check my other posts, I'm just a guy trying to make some money at sports until unemployment lets up...