Taking the OpenClaw Bitcoin Auto-Trading System One Step Further with Multi-Strategy Simulation#

2026-03-15

A few days ago, I wrote My First Personal Project with OpenClaw: A Bitcoin Auto-Trading System, where I explained why I built the first version of the system and how that initial structure worked.

If that post was a record of building the smallest version that could actually run, this post is about the next step. This time, instead of watching a single strategy in isolation, I expanded the system into a structure where multiple strategies are simulated side by side on the same market data and compared against each other.

I also saved the result of this stage as a multi-strategy simulation snapshot.

The core goal is simple: before connecting real money, I want to observe which strategies hold up better and react more consistently in different kinds of market conditions.

Note: This is not investment advice. It is a personal study note and project retrospective.

Why I switched to multi-strategy simulation#

The biggest thing I felt after building version 1 was that there is only so much I can learn by watching a single strategy.

For example, even if performance looks good on a certain day, it is hard to tell whether the strategy was genuinely strong or whether that day’s market simply happened to fit it well. The same is true when performance looks weak. It becomes difficult to judge whether the problem lies in the strategy itself or in the fact that the market regime was simply unfavorable for that kind of approach.

So I changed direction.

• I feed the same market data into multiple strategies at the same time.
• I let each strategy record its own independent paper-trading results.
• I compare performance at regular intervals and check how the rankings move.

This makes it much easier to see what kind of strategy is relatively favored by the market right now. More than individual buy or sell signals, what I wanted at this stage was an environment where I can observe and compare behavioral patterns across strategies.

What changed in this upgrade#

With this expansion, the system feels less like a simple auto-trading experiment and more like a small strategy lab.

1. I now run multiple strategies at the same time#

Including the original base strategy, I am now running six different viewpoints on the same stream of market data. Each strategy trades independently in simulation, without affecting the others. That makes it much easier to compare how differently they interpret the same price movement at the same moment.

2. I check comparison reports on a regular basis#

Previously, the main question was simply, “Did a signal fire?” Now I also care about, “Which strategy behaved more steadily?” I compare returns and trade results at regular intervals, and I also leave the full ranking in the daily summary.

3. I strengthened the market mood filter a bit#

Instead of looking only at price charts, I also added a way to take into account whether the market feels excessively overheated or excessively depressed. I would not call it sophisticated macro analysis, but it is enough to reflect at least one practical question: how conservatively should a signal be interpreted when overall sentiment becomes extreme?

4. I adjusted risk management to be more realistic#

I shortened the take-profit distance somewhat, and I added a mechanism that protects gains when price falls back after a profitable move. I did that because, in automated trading, limiting downside and protecting profit matter just as much as the strategy logic itself.

The strategies I am now running side by side#

The strategy documents I organized this time cover five strategies, and together with the existing base strategy they form the full comparison set. The short explanations below are written for non-developers, with the focus on a simpler question: what kind of market is each strategy trying to interpret, and how?

1. RSI oversold / overbought strategy#

This is the most intuitive one. If price seems to have fallen too much in a short period, it looks for a rebound. If price seems to have risen too much, it treats that as overheating and becomes cautious.

• Best fit: sideways markets with repeated swings but no strong direction
• Strength: simple and easy to understand, which makes it useful as a baseline
• Weakness: in strong uptrends or downtrends, it can enter against the move too early

2. MACD crossover strategy#

This strategy is more interested in moments when direction begins to change. It tries to detect whether upward momentum is coming alive or downward momentum is starting by watching how moving averages shift relative to each other.

• Best fit: markets with a relatively clear trend
• Strength: once it catches a trend, it can follow it in a fairly stable way
• Weakness: the signal tends to arrive a bit late, so it may miss the early part of the move

3. Bollinger Bands strategy#

This one focuses less on price level itself and more on volatility. It can look for mean reversion when price becomes statistically stretched, or try to catch the moment when a quiet market suddenly begins to move sharply.

• Best fit: rebound zones in sideways markets, or breakout phases after volatility compression
• Strength: useful for catching moments when the market suddenly comes alive
• Weakness: it can be vulnerable to false breakouts that quickly reverse

4. Golden cross / death cross strategy#

This is one of the most classic trend-following approaches. It watches for the point where a shorter average crosses a longer average, using that as a clue that the larger direction of the market may be changing.

• Best fit: markets where a medium-term trend continues with some persistence
• Strength: useful for trying not to miss the bigger move
• Weakness: it reacts slowly, so it can feel too sluggish for short-term trading

5. Triple-confirmation composite strategy#

This is the one I find most interesting right now. Instead of trusting a single indicator, it waits until different perspectives, overextension, trend, and volatility, line up before entering.

• Best fit: markets with many ambiguous signals where a more conservative approach is preferable
• Strength: it trades less often, but the logic becomes cleaner because it moves only when evidence overlaps
• Weakness: it can be so cautious that it enters late or misses good opportunities entirely

6. The original base strategy#

The original base strategy mixed overbought / oversold judgment with trend and momentum interpretation. In the multi-strategy simulation setup, I keep it as a kind of baseline so I can compare whether newer strategies are actually better or merely different in style.

I think this baseline matters a lot. It is not enough to keep adding new strategies. I also need a way to ask, is this genuinely better than what I was already doing?

The biggest advantage of this multi-strategy structure#

The biggest value of this upgrade is not that I have found the “correct” strategy. At least for now, it is almost the opposite. What matters is that I can now see much more clearly that different strategies fit different market conditions.

In sideways markets, a simple mean-reversion strategy may hold up better. In a strong trend, a trend-following strategy may have the advantage. A composite strategy may trade less often, but still shine by avoiding unnecessary entries. These differences are hard to see when everything is filtered through a single strategy, but they become much more visible in a multi-strategy simulation structure.

I think this is a necessary step before connecting live trading. In automated trading, what feels dangerous is not the mere fact that a signal appears. What feels dangerous is putting real money behind a signal without really understanding why I trust it.

What I want to watch next#

Rather than connect live trading immediately, I plan to monitor the system for another 1 to 2 weeks first. During that period, I especially want to watch the following:

• whether the performance gap between strategies remains somewhat consistent
• whether any strategy trades far too often or far too rarely
• whether the stop-loss and take-profit rules feel too sensitive or too blunt in real market movement
• whether strategy reactions become too aggressive in overheated or panicked market conditions
• whether the daily reports contain enough context to understand what actually happened

After that, I may move on to connecting the most stable-looking strategy to live trading. Even then, I would not start with large amounts. What I want is not one big win, but something closer to starting small, observing for a long time, and building a sense of operation that I can actually trust.

What this stage helped me realize#

What I confirmed again through this upgrade is that the essence of an auto-trading project is not just “writing one clever strategy.”

What matters more, at least to me, is building a structure where I can:

• compare strategies,
• leave results behind as records,
• and interpret those records afterward.

That is the kind of structure I want to build.

I do not want this project to become just a tool that buys and sells Bitcoin automatically. In the long run, I want it to become an experimental system through which I can slowly accumulate my own investing judgment. The move to multi-strategy simulation feels like one meaningful step closer to that goal.