Scalping Strategies 501: Strategy 89
The good news, Strategy 89 made $10K on one sim account in one day. The bad news, it also lost $1,200 when traded on another sim account using a different data feed. This happened all week. Why?
Important: There is no guarantee that ATS strategies will have the same performance in the future. I use backtests and forward tests to compare historical strategy performance. Backtests are based on historical data, not real-time data so the results shared are hypothetical, not real. Forward tests are based on live data, however, they use a simulated account. Any success I have with live trading is untypical. Trading futures is extremely risky. You should only use risk capital to fund live futures accounts and if you do trade live, be prepared to lose your entire account. There are no guarantees that any performance you see here will continue in the future. This post is for educational purposes only. I recommend using ATS strategies in simulated trading until you/we find the holy grail of trade strategy.
As a quick reminder, we’re on the hunt for the holy grail of automated trading strategies. If you have any questions, start with the FAQs and if you still have questions, feel free to reach out to me directly. If you have any trading based LLM ideas, applications or projects you would like to share, I’d love to hear them.
AutomatedTradingStrategies@protonmail.com
A list of all published strategies is available here.
This post has been a long time coming. Thank you Ashley for asking the right questions.
There's always been a bit of friction between technical analysts and scalpers in the trading community. Fundamentally, the rivalry stems from different trading philosophies. Technical analysts often focus on longer-term trends and patterns that develop over days, weeks, or even months. Scalpers operate on extremely short timeframes, within what many consider "market noise".
Scalping demands razor-sharp attention: surgical entries and exits. It’s close-range combat and can be extremely high-risk if you don’t know what you’re doing. Done right, however, it’s incredibly rewarding because you’re in and out quickly: inhale, exhale, profit.
Personally, I employ both technical analysis and scalping strategies in my manual trading—recognizing that different approaches can be effective in different market conditions. After all, you can’t catch a big wave unless there’s one to ride.
The Heavy Lifting: Forecasting, Slippage, and Missed Trades
All things being equal, the holy grail of automated trading strategies is more likely to be a scalping strategy than a trend-based, long-range one. The scalping path is always ≥ the trend path. The key insight is that capturing multiple small moves can add up to more than a single larger move over the same price range. Put simply, the zigzag path covers more distance than the straight path. The ratio increases with oscillation frequency and amplitude, which is why I like scalping on NQ—higher volatility leads to more pronounced zigzag patterns and a greater difference in path lengths.
Of course, the phrase "all other things equal" is doing a lot of heavy lifting, especially when it comes to automating the process. Capturing every pivot point is tough, and increased slippage, missed trades, and inflated transaction costs can weigh down any strategy. Scalping in particular demands deep understanding of execution limitations. In general, the longer you plan on holding a trade, the less important missed trades and/or slippage is.
"Missed trades" occur when your order isn’t filled at all, often due to low liquidity or large price gaps. Unlike slippage—where you get filled at a different price—no execution means you’re left out entirely. I had to learn this the hard way after being impaled by our first unicorn: Strategy 10. In longer-term trading, these issues matter less; in scalping, they’re critical, especially if your win rate is below 60%.
In addition to slippage and missed trades, consider the likelihood that your performance results could be completely off due to how the backtest is calculated. In NinjaTrader 8 (NT8), there are three different options for bar calculation:
On each tick
On price change
On bar close
"On each tick" and "on price change" require more computation and can be prone to latency issues, but they’re better for the kind of surgical entry required for scalping strategies. This often translates into high slippage at best, and at worst, wildly different performance results. Call it the Heisenberg principle of trading: the more precisely you try to measure, the more you skew the outcome.
Scalping Strategies: Automation Challenges
Naturally, automating this process is very hard to do. No backtest can be fully trusted, especially not tick-by-tick. Even forward tests, which are closer to reality, can fail to predict live results. So, as we refine a framework for automated strategy development in general (see Strategy 88), we need a separate sandbox for scalping strategies. We need a space that allows for more collaborative and rigorous testing. In traditional systems, the challenge is developing a strategy that works across all market regimes. In scalping, the challenge is figuring out how to test these strategies because traditional approaches don’t cut it. I’ve lost plenty of money live-testing scalping strategies that performed well in forward tests, but failed in live conditions. I consider it a tuition fee for the market’s higher education, but it doesn’t mean I like paying it.
What’s an automated scalper to do?
The first step is a forward test. There’s no getting around it for strategies that calculate on each tick. I’m currently forward-testing about 20 scalping strategies and roughly 60 variations on these strategies. I haven’t shared any of them with you yet, but I will share the best performing scalping strategy in the forward test with you today. My hope is to prune this list down to ten variations by the end of the month.
Then what? Just because these ten perform well in a forward test doesn’t mean they’ll deliver the same results in a live account. Price differences, calculation errors, order management—plus a host of other order flow issues—can change the game once real money’s on the line.
What else can we try before running scalping strategies live?
Try a dedicated server: I’m currently using a virtual server, but scalping strategies can be computationally intensive. Maybe I need to throw more dedicated resources at the problem.
Try additional data sources: I primarily use one broker. Perhaps I should incorporate multiple price feeds to diversify the data.
Funded Trader Programs: While I’m not their biggest fan, these programs can be a quick way to test scalping strategies in another environment. (Thanks for the suggestion, Cora.)
Micro Contracts: I usually only deploy micros on the pullback, but this could be another way to test scalping strategies without exposing myself to full risk.
There are also methods for structuring scalping strategies to reduce issues with order acceptance, such as specific use of limit orders. I’ll maintain a living list of best practices at the bottom of this post.
Finally, I’m a sucker for all things LLM, so I took OpenAI’s early Christmas present for a spin. For $200, I get the privilege of tinkering with what’s touted as the most intelligent AI on the market. While Claude is currently my LLM of choice, I have no real loyalties. If o1 is better, I’ll use it. My plan this weekend is to create a few scalping strategies using o1 and get them into the forward test for next week. With the help of LLMs, maybe we can develop a few techniques that yield better scalping strategies. If I find anything that outperforms Strategy 89, I’ll let you know.
Oh, these are exciting times for traders—market ninjas, code junkies, and data devotees alike. Let’s see if we can’t carve out an edge together.
Strategy 89: The Contrarian
The strategy I’m going to share with you today is the highest performing scalping strategy so far in the Scalpers Forward Test (not to be confused with the Q4 Forward Test). Again, all forward tests use a simulated account on live data.
I’ve only been forward testing this strategy for one week, but it has performed well every day, making between $1K and $10K per day. Click here for the real-time trades from December 5-6 (note, net profit does not account for any transaction costs, which could be significant).
Here’s a summary analysis of performance.
BEFORE you get too excited, I also tested the strategy simultaneously using another simulated account on different live data feed and it FAILED miserably. Instead of up $10K, it ended the day down close to $1,200. Naturally, I am reluctant to run this on a live brokerage until I understand what’s going on, but I wanted to share this with you so you can troubleshoot yourself.
I think there’s something here. We just need to clean it up a bit.
Does anyone have any ideas about why this is happening? Whatever your experience is with Strategy 89, I would like to hear it. Personally, I think the issue is a function of many things and I’ll share those things with you as we continue to test the strategy next week. I’ve created a sandbox to house the conversation here, in ATS Chat. If we need a more collaborative space, I’ll migrate the conversation to Slack.
Okay, now let me share the strategy description with you. The download link is at the bottom of the post.