Interpreting Your Results
Beyond the Numbers
After every trading session, LLMTrader provides a comprehensive analytics dashboard showing how your AI-driven strategy performed. This guide explains what each metric means, how to read the charts, and how to distinguish genuinely strong performance from misleading results.
Understanding your results is not optional; it is the foundation of improvement. The users at the top of the leaderboard are not just good at writing prompts. They are good at reading results and turning those insights into better prompts.
Key Performance Metrics
Total Return (%)
What it is: The percentage change in your portfolio value from session start to session end.
How to read it: A session that starts with $10,000 and ends at $11,200 has a total return of +12%. A session ending at $9,500 has a total return of -5%.
What to watch for: Total return is the most visible number, but it is also the least informative in isolation. A +30% return with a 40% drawdown is very different from a +30% return with an 8% drawdown. Always read total return alongside risk metrics.
Sharpe Ratio
What it is: A measure of risk-adjusted return. It tells you how much return you earned per unit of risk (volatility) taken.
How to read it:
- Below 0.5: Poor risk-adjusted performance. Returns did not justify the volatility.
- 0.5 to 1.0: Acceptable. Reasonable returns for the risk taken.
- 1.0 to 2.0: Good. Strong returns relative to the volatility experienced.
- Above 2.0: Excellent. High returns with relatively low volatility.
What to watch for: Sharpe ratio can be inflated by very short sessions or calm market conditions. A high Sharpe over 3 trades is far less meaningful than a high Sharpe over 50 trades. Consider it alongside the number of trades and the duration of the session.
Maximum Drawdown (%)
What it is: The largest peak-to-trough decline in portfolio value during the session. It answers the question: "What was the worst period I had to endure?"
How to read it: A maximum drawdown of -15% means that at some point during the session, your portfolio fell 15% from its highest value up to that point.
What to watch for: Maximum drawdown is the single most important risk metric. A strategy with a -5% maximum drawdown is fundamentally different from one with a -30% maximum drawdown, even if both end with the same total return. Ask yourself: "Could I have held through this drawdown without panicking?" If the answer is no, the strategy is too aggressive for you.
Calmar Ratio
What it is: Total return divided by maximum drawdown. It measures how much return you earned for the worst pain you experienced.
How to read it:
- Below 0.5: The drawdown was disproportionately large relative to the return.
- 0.5 to 1.0: Acceptable. The return reasonably compensated for the drawdown risk.
- 1.0 to 2.0: Good. Strong returns relative to the worst drawdown.
- Above 2.0: Excellent. High return with controlled drawdowns.
What to watch for: Calmar ratio is especially useful for comparing strategies with very different risk profiles. It normalizes performance by the worst-case experience.
Win Rate (%)
What it is: The percentage of trades that were profitable.
How to read it: A win rate of 60% means 6 out of every 10 trades made money.
What to watch for: Win rate alone is deeply misleading. A strategy can have a 90% win rate and still lose money if the losses are much larger than the wins. Conversely, many successful strategies have win rates below 50%; they lose often but win big. Always pair win rate with average trade P&L.
Average Trade P&L
What it is: The average profit or loss per trade, usually expressed as a percentage or dollar amount.
How to read it: An average trade P&L of +0.8% means that across all trades, the average result was a 0.8% gain.
What to watch for: Look at this alongside win rate. The combination tells the full story:
- High win rate + small average P&L = Many small wins; strategy works through consistency
- Low win rate + large average P&L = Few big wins carry the strategy; most trades lose small
- High win rate + large average P&L = Exceptional (and rare) performance
- Low win rate + small average P&L = The strategy is not working
Reading the Charts
Equity Curve
The equity curve plots your portfolio value over time. It is the most intuitive visual representation of session performance.
What a healthy equity curve looks like:
- Generally trending upward (or at least not downward)
- Relatively smooth, without extreme spikes or crashes
- Consistent slope, not dependent on a single explosive move
Red flags in the equity curve:
- A single massive spike that accounts for most of the return. This suggests luck, not skill.
- A long, steady decline punctuated by a recovery at the end. The final P&L looks okay but the journey was terrible.
- Wild oscillations with no clear trend. The strategy has no edge and is essentially random.
Drawdown Chart
The drawdown chart shows the percentage decline from the running peak at each point in time. It is always at or below zero.
What to look for:
- Depth: How deep do the drawdowns go? Deeper means more pain.
- Duration: How long do drawdowns last before recovery? Longer recovery periods are more difficult to endure.
- Frequency: Are drawdowns rare and sharp, or frequent and moderate? This tells you about the strategy's risk character.
A strategy with frequent shallow drawdowns (2-3%) that recover quickly is very different from one with rare deep drawdowns (15-20%) that take a long time to recover.
Trade Distribution
The trade distribution chart shows the spread of individual trade results, typically as a histogram.
What to look for:
- Center: Is the distribution centered above or below zero? Above zero means the average trade is profitable.
- Spread: How wide is the distribution? A tight cluster around a small positive number suggests a consistent strategy. A wide spread suggests high variance.
- Tails: Are there extreme outliers? A single massive winner or loser can dominate overall results and may not be repeatable.
A well-functioning strategy typically shows a distribution that is slightly right-shifted (centered above zero) without extreme outliers dominating the results.
Understanding the Trade History Table
Every trade in your session is logged with detailed information. Here is what each field means.
| Field | Description |
|---|---|
| Timestamp | When the trade was executed. Use this to correlate trades with market events. |
| Asset | Which cryptocurrency or trading pair was traded. |
| Action | Buy, sell, or close. Indicates what the AI did. |
| Size | The position size, either as a percentage of portfolio or an absolute amount. |
| Entry Price | The price at which the position was opened. |
| Exit Price | The price at which the position was closed (for completed trades). |
| P&L | Profit or loss on the trade, shown as both a dollar amount and a percentage. |
| AI Reasoning | The model's explanation of why it made this decision. This is one of the most valuable fields; read it carefully. |
How to Use the Trade History
Identify patterns. Are losses concentrated in a specific asset? During a specific time of day? In a particular market condition? Patterns in your losses tell you what to fix.
Read the reasoning. The AI reasoning field reveals whether the model understood and followed your prompt. If the reasoning does not match your intended strategy, your prompt needs refinement.
Check sizing discipline. Did the model respect your position sizing rules? Inconsistent sizing is a common issue that can be addressed with more explicit prompt language.
Look for revenge trading. Did the model take oversized positions immediately after a loss, trying to recover? This is a behavior pattern to watch for and explicitly prohibit in your prompt if you see it.
Red Flags: When Good Numbers Lie
Huge Returns with Huge Drawdowns
A +50% return sounds amazing until you learn the session hit -35% along the way. This is not a reliable strategy; it is a risky strategy that happened to recover. Next time, it might not.
What to do: Look at the Calmar ratio. If it is below 1.0 despite a large total return, the strategy is taking too much risk for the reward.
Single Lucky Trade Masking Losses
Sometimes a session shows a positive total return, but the trade history reveals that one enormous winner is masking a string of losses. Remove that one trade and the strategy is deeply negative.
What to do: Check the trade distribution. If one trade is a massive outlier, ask yourself: "Was that repeatable, or was it luck?" If you cannot confidently say it will happen again, the strategy needs work.
High Win Rate, Negative Returns
A 75% win rate feels good, but if the average loss is 4x the average win, you are losing money despite winning most trades. This is the classic "picking up pennies in front of a steamroller" pattern.
What to do: Compare average win size to average loss size. If the ratio is unfavorable, adjust your prompt to tighten stop losses or let winners run longer.
Incredible Sharpe in a Short Session
A Sharpe ratio of 5.0 over 10 trades is statistically meaningless. The sample size is too small to draw conclusions. This number will almost certainly decline toward reality over more trades.
What to do: Be skeptical of any metric calculated from fewer than 20-30 trades. Extend session duration or run multiple sessions to get a meaningful sample.
Flat Equity Curve that Suddenly Spikes
If the equity curve is flat for 90% of the session and then surges at the end, the strategy did not have a persistent edge. It sat around doing nothing and then got lucky (or the market moved sharply in its favor). There is no evidence the strategy itself is valuable.
What to do: Look for strategies that produce consistent, steady returns over time. An upward-sloping equity curve with moderate variance is far more trustworthy than a hockey stick.
Putting It All Together
The strongest session results show a consistent pattern across all metrics:
- Positive total return
- Sharpe ratio above 1.0
- Maximum drawdown that is tolerable and proportional to returns (Calmar above 1.0)
- Win rate that makes sense for the strategy type
- Average trade P&L that is positive
- An equity curve with a steady upward slope
- A trade distribution centered above zero without extreme outliers
- AI reasoning that aligns with the intended strategy
When most or all of these boxes are checked, you have found a strategy and model combination worth refining further. When several are off, it is time to revisit your prompt, try a different model, or reconsider your approach entirely.
Remember: One great session does not prove a strategy works, and one bad session does not prove it fails. Look for consistency across multiple sessions and different market conditions. That is where real confidence comes from.