A slightly losing/winning system is the worst. Why? Because it is a break even system basicly. A consistently profitable system is making money. A consistently losing system can be reversed. Consistency is the key. A consistently breakeven system sucks, because it is neither making money nor can be reversed to make money. So I suggest you try to make a new system... Edit: On second thought, there might be a way. Check out the trades and see how many winners follows how many losers. If you see a pattern, wait for a bunch of losers, then increase size for a short period...
Why do you think the 6E did so much better? Because it's "trendier"? Is that over the 10 year testing period? How difficult is it to change parameters in those tests?
I'm a Kelly trader, and the fact is that Kelly sizing will not turn a losing system into a winning system. If a trader can't make a system profitable with a fixed trade size, there is no fractional sizing scheme that will do the trick. So says the math. So the title of this thread is hopeless. "Almost profitable" means break-even, and Kelly says you always bet zero on a break-even system.
Disagree. I am profitable because I average into trades. A fixed sizing scheme would not work for me, unless I misunderstood what you are saying.
Except there is a huge difference between a breakeven and a losing system. The losing system, you want to reverse, the breakeven system, you search for patterns and size accordingly.... Now of course we assume here that you can see some sort of repeating pattern in the win/loss....
If it was as simple as "reversing" a losing system to turn it into a winner, we'd all be rich. You can't reverse commissions and slippages, for starters. And breakeven systems cannot be made profitable. Consider the simple coin flip. This is a breakeven system IF you use a fixed bet size. BUT ... if you use a fractional bet size, it becomes a losing system. Let's say you decide to bet a constant 5% of your account on the coin flip. So half the time your account increases to 105% and the other half it decreases to 95%. So on average your account decreases to sqrt[ 1.05*.95 ] = 99.875% per bet. This is why Kelly says your best bet for breakeven systems is zero.
Partially right. Depends on why the system is a loser. If it is just because commissions, right, we can't reverse it. But if it has consistent big losses, reverse away!! And breakeven systems cannot be made profitable. Consider the simple coin flip. This is a breakeven system IF you use a fixed bet size. But we don't use a fix sized bet, we increase when we expect winners. Your example later on didn't explain anything, I am not sure why you included it... Here is a win (1) loss(0) result of a breakeven system: 1,1,0,1,1,0,0,0,1,0,1,1,1,0,0,1,0,0,0,1,0,1,1,0 etc.etc. Now the pattern I see here is that seldom there are 4 losers in a row. So if the OP has a similar system, after 3 losers in a row, I would go heavier... It might not be the best solution, but that is my best answer how to make a breakeven system profitable...
Futures trading doesn't always involve taking profit in percentage. Half the time you lose 5 points. Half the time you gain 5 points. You're back where you started from (not including commissions).
1a2: After I dig into the thread mentioned earlier for the original criteria, I can run it through a couple optimizations. Currently I am testing snowball momentum plays but using some heavier MAs as the ignition (specifically two crossovers, one being the 250-300). By looking at the speed of the move first, what clues are there to the strength of this trend that is forming. That is the filter. Don't want chopping or early trends, but the moment a trend is signaling a bigger bet. Ever had that thought on the day's chart: is this day going to trend or revert to mean? Usually that question comes up because you are seeing signs of an early established trend. If momentum stalls it is going to rollover big. If it keeps going you've got a nice break and your entry. (This n-o-t about gathering data from something like the RSI momo indicator.) Maybe none of this is helpful, I don't do a lot of strat testing on daily charts. Only mid (1H) to smaller frames. Drift of two MAs and back to the mean is another one of the list to explore. Last week was Kirshenbaum It only takes a week or less of hacking away to figure out if you have something worthwhile or not.