Measuring Win Rate for a Trend-Following Strategy
Summary
The document asks how to calculate a basic win percentage for a trend-following model that produces long and short signals. The proposed approach is to compare the price when a signal begins with the price when the signal changes, count a trade as a win if the price moved in the signal's direction, and divide wins by the number of signals.
The excerpt contains only the question and an illustrative price table; it provides no answer, backtest results, or evidence that this measure is sufficient. The suggested calculation raises unresolved choices such as how to define trade entry and exit prices, account for short positions, handle signals that persist or overlap, and include costs. A raw win rate also does not measure the size of gains and losses, so it cannot alone establish whether the strategy is profitable or useful.
Key ideas
- The proposed win rate counts signal intervals with price movement in the position’s direction.
- A consistent definition of when a signal starts and ends is needed to calculate trade outcomes.
- Long and short positions require direction-aware return comparisons.
- Win rate alone omits the size of gains and losses, as well as trading costs.
Tags
Full text
# Calculating basic win% of simple trend-following strategy?
# Calculating basic win% of simple trend-following strategy?
I apologize if this isn't the correct place to post this. I'm not quite sure where else I should post on stackexchange.
I'd like to preface this by saying I'm not actually planning on trading. This is more of a thought experiment carried out in my own free time. I have a very basic trend-following model that looks at several technical indicators. The model returns signals that indicate whether to take a long (1) or short position (-1). I'm interested in calculating some sort of simple win% to see how many times this model has potentially made a good trade signal.
I'm not quite sure how to go about this. For instance, I have something like
```
Position High Low Close Open
`1 3.375 3.3545 3.375 3.355 3.3985
0
0
0
1 3.4075 3.3935 3.397 3.4065 3.4134
```
Would I then look at prices (3.375 close) at the first signal and prices when at the signal changed (3.3935 close)? Since prices increase on a long signal, then this would count as a win. Let's call this $W_1$. If I repeated the same calculation throughout my series and got the number of wins $\sum_{i=1}^{n} W_i$, would it then make sense to divide this by number of long signals to get a win% $\frac{\sum_{i=1}^{n} W_i}{\text{# long signals}} $.
This seems a little silly and naive to me, so I'm open to feedback!Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.