Have you analyzed your systems, to see if the expectation really is the same after a winning trade, and a loosing one?

If the trading results are a stochastic process (due to market being another stochastic process), there may be a correlation from earlier trading result(s) and the next one?

I have a system where it seems to be that the expectation is not the same.

6 years backtest, and 2 years live.

I dont know (and dont care) what causes the deviation.

But, if i speculate:

The system is trend following.
After a trend, where the system makes money, there is in many cases a consolidation, where the system looses money.

Therefore it may be that after a winner ("trend ends"), the next expectation is less ("consolidation / range / flag etc".)

I have another system which seems to work the opposite, i.e. the expectation seems to be negative after the first loosing trade, and positive after a winning trade.

You do not know, before you study your system?

Otherwise, you only set an assumption or hypothesis, that trades are fully independent.

2tenor 

Have you analyzed your systems, to see if the expectation really is the same after a winning trade, and a loosing one? If the trading results are a stochastic process (due to market being another stochastic process), there may be a correlation from earlier trading result(s) and the next one? I have a system where it seems to be that the expectation is not the same. 6 years backtest, and 2 years live. I dont know (and dont care) what causes the deviation. But, if i speculate: The system is trend following. After a trend, where the system makes money, there is in many cases a consolidation, where the system looses money. Therefore it may be that after a winner ("trend ends"), the next expectation is less ("consolidation / range / flag etc".) I have another system which seems to work the opposite, i.e. the expectation seems to be negative after the first loosing trade, and positive after a winning trade. You do not know, before you study your system? Otherwise, you only set an assumption or hypothesis, that trades are fully independent.

I mean it sounds about right if you're leaving out the context of the market. I've come to a lot of quaky conclusions doing back tests over long periods of time and even manually trading data from random periods after it has transpired. What I've realized is that the decision making process for trading is contingent upon market internals such as sentiment, day of the week, news and reports, and most of all, the larger time frame.

Without context what you're going to find is that a trend usually trends until something prevents it from trending (news or a report). A market will range until something forces it to trend. And, lastly, a market will trend and range and vise versa when hit with multiple data inputs during a short period of time.