September was the bot's sixth live month and a month of two halves. The first two weeks were weak, the last three were solid, and the main strategy finished above break-even before charges. Here's the month from the trade log.
At a glance
- The main strategy finished above break-even. It placed 27 of 43 live trades (63%), won 51.9% of them, and its wins outweighed its losses before charges: a profit factor of 1.13. It is a small sample, so I treat it as encouraging rather than proven.
- The month got better as it went. The first 12 live records, to 13 September, won 4 and had a profit factor of 0.27. The 31 records from 14 September won 19 and had a profit factor of 1.47.
- Losses stayed small. No live loss was bigger than 2.5 average losses, and the best live win was 2.8. The largest main-strategy loss was 1.7.
- Over half of live trades won. 23 of 43 (53.5%), with a longest run of 6 winners in a row.
- Paper was roughly flat. The main strategy on paper won 48.7% of 583 records with a profit factor of 1.01. That is close to break-even, so it neither confirms nor contradicts the live result.
- The other live strategies cost the month. Their 16 records had a profit factor of 0.58, which kept all live trading at 0.81.
The scorecard
| Live | Paper | |
|---|---|---|
| Trade records | 43 | 4,018 |
| Trading days with activity | 13 | 20 |
| Strategy variants | 8 | 71 |
| Win rate | 53.5% | 43.6% |
| Payoff ratio (average win ÷ average loss) | 0.71 | 1.23 |
| Profit factor (all wins ÷ all losses) | 0.81 | 0.99 |
| Longest winning streak | 6 in a row | 21 in a row |
| Longest losing streak | 5 in a row | 31 in a row |
| Deepest drawdown (in average losses) | 8.8 | 206.7 |
| Ran to the bot's own exit | 26 of 43 (60.5%) | 3,989 of 4,018 (99.3%) |
Live and paper are always kept apart. Paper runs many more ideas, including ones that never go live, so it is the wider and noisier of the two.
Where the result came from
The main strategy placed 27 of the 43 live trades. Results below are in units of the month's average live loss, so a figure of +1.0 means the group made back one average losing trade, before charges.
| Live records | Share | Win rate | Profit factor | Net result (in average losses) | |
|---|---|---|---|---|---|
| Main strategy | 27 | 63% | 51.9% | 1.13 | +1.1 |
| Other live strategies | 16 | 37% | 56.2% | 0.58 | -4.8 |
| All live trading | 43 | 100% | 53.5% | 0.81 | -3.7 |
The single best live record was worth 2.8 average losses and the single worst 2.5.
What worked
The main strategy held its shape. Its largest loss was 1.7 average losses and its best win was 2.8. It won 14 of its 27 live records and stayed above break-even before charges, at 1.13.
The second half of the month was steady. From 14 September, three weeks in a row finished with a profit factor above 1 (1.65, 1.47 and 1.42), on 5, 14 and 12 live records. The weeks are small, so I'd call this a better run, not a new normal.
Losses were controlled. 20 of the 43 live records were losses, and none was larger than 2.5 average losses. The average winning trade was 0.71 of the average loss, which is not enough on its own, but the absence of large losses kept the damage limited.
What held it back
Too many trades were closed by hand. 17 of the 43 live records (40%) were closed by hand, 15 of them on the main strategy, all in the last two weeks (5 in the week of 21 September, 12 in the week of 28 September). The bar I set in May was five or fewer. Over this small sample the by-hand exits did better than the rule-based ones (profit factor 1.97 on 17 records, against 0.47 on 26). I don't read that as a reason to keep doing it. It tells me the rules' exits need work.
The payoff ratio is still below what the win rate needs. The average live win was 0.71 of the average live loss. At a 53.5% win rate, break-even needs about 0.87.
The other live strategies were a drag. Eight strategy variants went live, and the 16 records outside the main strategy had a win rate of 56.2% but a profit factor of 0.58.
The first two weeks were weak. Only 4 of the first 12 live records won. The month's profit factor of 0.81 is before charges, so after brokerage and taxes the live result was lower still.
How trades ended
| How the trade ended | Live | Share of live | Share of paper |
|---|---|---|---|
| Closed by hand | 17 | 40% | 1% |
| Trailing stop | 12 | 28% | 16% |
| Stop-loss | 8 | 19% | 17% |
| Other rule-based exit | 4 | 9% | 53% |
| Target reached | 2 | 5% | 5% |
| Partial profit booked | 0 | 0% | 4% |
| Time limit or end of day | 0 | 0% | 5% |
What I'm watching in October
- Manual exits back to five or fewer. September's 17 of 43 is well above that bar, and I'd rather fix the rules than keep stepping in.
- Payoff ratio above 0.87 at this win rate. That is what turns a slim edge into a clear one.
- Whether the main strategy's 1.13 holds over more than 27 live records, and whether paper (1.01 over 583 records) moves closer to it.
- The other live strategies' record. A profit factor of 0.58 over 16 records is too small to settle anything, but it needs to improve.
- Incomplete data to note. Live trading logged 43 records on 13 trading days, and the last live record was on 29 September. Paper logged 4,018 records on 20 days. With this few live trades, I'm not drawing conclusions about time of day or expiry days.
Every scorecard uses the same layout and definitions, explained on the scorecard page. Figures are counts and ratios from the trade log, before brokerage and taxes. For how I read the log, see how I analyse my trading bot's trades.
This scorecard describes my own trading system, for information only. It is not investment advice, and past results say nothing certain about the future. I am not registered with SEBI as an investment adviser or research analyst.