Peter - Cracking Markets

I’ve been trading for ~30 years. First half: fully discretionary, living inside futures microstructure. It worked—until algos started exploiting the same patterns and reacting in microseconds. Edge decay was real. So ~10 years ago I switched to systematic. Now I run many uncorrelated strategies in parallel without babysitting screens all day. I wouldn’t go back. My biggest unlock: reusability of know‐how. When I finish a new system, I plug it into a ready workflow in minutes. It monitors itself; I move my brain to the next big thing. Here’s the playbook I wish I had from day one: - Framework (design once → reuse forever) - Data → clean, feature, label. - Hypothesis → simple, testable edges (breakouts, momentum, mean reversion). - Validation → IS/OOS, realistic costs/slippage. - Risk → position sizing, max heat, portfolio exposure caps. - Deploy → automated orders, fail‐safes. - Monitor → health dashboards, kill‐switch rules, mobile app. - Iterate → new systems slot into the same pipeline. Principles that compound: - Many small, independent edges > one “genius” setup. - Process beats prediction. - Shipping beats perfecting. Discretionary taught me markets. Systems gave me scale.

This improved my LLM-written trading tools more than any “better prompt”: I stopped trusting generated code. And started forcing it through an audit protocol. LLMs can write decent code today. Sometimes very good code. But trading automation is not just: “Does the function run?” It is: “Can this survive execution, broker state, account scope, order ownership, partial fills, restarts, stale orders, and live exposure?” The most dangerous bugs are not syntax bugs. They are quiet assumptions. Examples: - position read from the wrong account - protective stop missing after restart - cancel request treated as cancellation confirmation - partial fill handled as if the order was fully cancelled - stale stop left live after the strategy is flat - order without a unique ownership marker - backtest rule that cannot be reproduced live So I do not ask the LLM vague questions like: “Is the risk management okay?” That is almost useless. I create gates. Concrete checks the system must prove. Example: “Every live long position has exactly one valid protective SELL stop or one owned closing SELL.” Then the LLM must inspect the code and answer each gate: PASS / PROBLEM / NOT PROVEN The audit document is not frozen. It becomes a living protocol. Every bug, edge case, broker issue, and production surprise becomes a new gate I use to test all my code - past and future.

Working note from my automated long 0DTE breakout option bot: This is not the final version yet. The current bot buys ITM 0DTE options based on my volatility breakout signals. Execution is still very basic - MKT orders at the breakout - so slippage is painful. But it does not trade every day. On the contrary, it usually takes only a few trades per month. And yet, since inception: Bot: +58.5% SPY: +41.7% This is running on a smaller account and represents only the first stage of the full 0DTE automation project. Why did I start with long options? Because it was the easiest module to implement first with my current time limits. The full bot should also sell options for premium, but before combining both sides, I wanted to validate whether systematic 0DTE buying has real potential when it is tied to volatility expansion. The main lesson so far: 0DTE buying is not about being active every day. It is about waiting for the few sessions where intraday movement is large enough to overcome spread, decay, slippage, and execution costs. Most trades will lose. The equity curve will be volatile. But when the breakout is real, convexity pays. My takeaway: There is real potential in systematic 0DTE buying on SPY/QQQ when it is driven by volatility breakout logic, not random directional guessing. After validation, SPX/NDX may be even cleaner because they are cash settled. Current plan: - improve execution with limit chasing around mid instead of raw MKT orders - finish the premium selling module - solve quote database handling, because the options quote data size is currently my bottleneck

@Phille94 Yes, I do. My account is based with IBKR EU as well. You just need professional status to trade ETFs.

@SystematicPeter Do you trade us etfs on the intraday breakout? I also live in Europe, and us ETFs are off limits on IBKR for me.

I don’t trade one momentum strategy. I trade momentum as a portfolio across multiple timeframes. In this video I show my live IBKR platform and explain how I combine monthly rotation, swing breakouts and intraday momentum using systematic rules, diversification and automation.

@DVTrader1 This I do manually for some strategies. I do have a separate automated dashboard in Python, but I do not think you can separate the strategies like this via API.

@SystematicPeter This is awesome! How are you able to separate your strategies in TWS view like that? I'm currently relying on tagging my trades thru the IBKR API with the "orderRef" field and then using a custom reporting tool

@retire_readyinv I am currently using Claude Code for coding and GPT for review. It is interesting that so many people mention Grok. I have tried working with it several times, but I just did not find it better than GPT.

@SystematicPeter Great points. I use Gemini and Grok for volume work but ultimately feed that into Claude for final analysis.

The weakest way to use LLMs in the markets is to say: “Build me a profitable trading strategy.” That is how you get hallucinations. If you want an LLM to become a serious partner in systematic trading, you need to give it verifiable context and a detailed implementation plan. The real work is not prompting. The real work is planning. I often create this plan together with an LLM, but it is not a passive process. You need to think, challenge, refine, and verify every step. One useful trick: Use multiple LLM models to review the plan. They have different “personalities” and often catch different problems. In trading automation, vague input does not just create bad output. It can create expensive output. Detailed plan + verifiable context = a much more useful trading assistant.

Most European IBKR traders don’t realize this: Even as retail clients, they often have access to much lower stock margin requirements than U.S. retail traders. For large, liquid U.S. stocks, IBKR Europe commonly uses risk-based margining. In practice, maintenance margin can be around 15-25% - and this is not limited to intraday trading. It can also apply overnight. With Professional Client status, margin requirements on ETFs like SPY or QQQ can even be below 10%. Does this mean you should run your full account on maximum leverage? Absolutely not. Leverage is where many traders destroy themselves. But in systematic portfolio trading, temporary access to higher buying power can be very useful. Not to gamble bigger. But to handle periods when more independent systems are active at the same time, or when trading higher-priced stocks/ETFs with smaller profit targets and defined stop-loss levels.

Live Trading Models - Intraday Volatility Breakout: crackingmarkets.com/intraday-volat… Detailed Statistics (Updated Daily)

New equity high on my intraday volatility breakout strategy in Nasdaq. Nothing complex here. It is simple bracket trading: - first define when to trade - trading every day works, but commissions and slippage can eat too much of the edge - my results include fees - use the open as the reference price - place long trigger at Open + ATR - place short trigger at Open - ATR - use a small protective stop - exit at end of day if the stop is not hit That last rule matters. Many traders kill these systems by taking profits too early. The edge is not in predicting the direction. It is in letting intraday volatility expand after the breakout triggers. I also track this as a simple portfolio across S&P 500, Nasdaq 100, and Bitcoin futures, with daily updates on my blog. There is also a deeper breakdown with sample code for traders who want to study, test, and adapt the idea properly. Simple rules. Clear execution via automated bracket tools. No discretionary guessing.

@EutheniaADV Some strategies are just small variants of others, so I have about 15 strategies, but not all of them are really unique. Some are the same strategy trading different markets, etc. I definitely want to add more unique alphas.

@SystematicPeter Nice! How many live strategies do you have running?

Working note from my live Nasdaq momentum book: In mid-April I shared backtests here, including the rules: Long Donchian Channel breakout + trailing stop on high momentum Nasdaq stocks. I have now started running the model live on selected high momentum names. The selection is purely mechanical. I liked the idea on its own, but I like it even more now as a complement to my NDX relative momentum system. The reason is practical. My NDX momo model is robust, but it holds positions for a month. That monthly hold is part of the edge. But it also creates a very specific sizing problem: the system has to survive a large drop in an individual stock before the next rebalance. The Donchian setup has a different job. - It waits for the breakout. - It follows price with a trailing stop. - It can still suffer from an overnight gap. - But in normal conditions it should exit on the first real pullback. The trade-off is that it is more expensive to operate. When markets chop and breakouts do not follow through, this type of model pays for a lot of failed attempts. But right now, NDX names are moving a lot. Breakouts are clustering. And many stocks are starting to behave almost parabolically. For me, this is the kind of tape where swing breakout models deserve more attention. MU is a good current example. NDX momo entered on the open at the beginning of the month using the ranking model and will hold for the monthly cycle. The Donchian breakout entered later, on the actual breakout, and will exit using trailing stop-loss logic. Not every Donchian breakout overlaps with my NDX momo positions. But when these setups appear, they often point to the strongest plays in the market. And this is exactly why I do not want to trade these moves with fixed-hold logic only. What goes up fast can also come down fast. So I prefer to participate with a trailing stop. I am not trying to guess how far MU can go. I only need a rule that keeps me in while the move is working - and forces me out when it stops.

@systematicls Exactly 😂

@SystematicPeter x.com/systematicls/s…

It's so interesting to me how drastically different the personalities of Opus 4.7 and GPT 5.5 are. GPT 5.5 is smart, hardworking, slightly autistic and doesn't want to talk much. Huge preference for diving into tasks and hammering tasks out over long, long periods. Opus 4.7 is also obviously smart, but the only way to accurately describe Opus 4.7 is that he is charismatic but extremely lazy, talks a lot, but doesn't like to get work done. The only saving grace of Opus 4.7 is that he really does have an eye for design. -- Opus 4.7 is the quintessential toxic employee you need to fire while GPT 5.5 is the one you're glad to have on the team, but wished you could understand him better / have a more pleasant time interacting with him.

@WolfhartInvest I do not have IV data going that far back in history. But you can backtest any kind of mean reversion yourself.

@SystematicPeter would love to see the blue line from 1995 on!

I used to judge trading strategies one by one. That was the mistake. The real improvement came when I stopped looking for the perfect model and started combining simple edges that behave differently. A minimal systematic portfolio I would start with: US momentum + US long dip stocks. In the chart: - Green = NDX Momentum - Blue = Deep Dip - long mean reversion - Black = both strategies traded together - Gray = NDX benchmark These are two models from my Live Trading Models blog section, with positions updated daily. Simulation rules: - never more than 100% invested - no leverage - fees included Just these 2 strategies together: - CAGR: 24.5% - Max drawdown: -16% - Sharpe: 1.64 - Average capital usage: 42% That last number is the real key. 42% average capital usage means the portfolio is not fully loaded all the time. So you still have room to overlay other systematic edges: - fast intraday volatility breakout - option premium selling - momentum on other markets - other long and short models This is where systematic trading becomes powerful for retail traders. Not because one model is perfect. But because different models solve different market regimes. One catches momentum. One buys panic. One trades intraday volatility. One collects premium. Simple strategies. Different return drivers. Controlled exposure. Automation where possible. It still requires work. Testing. Execution. Monitoring. Risk control. Position sizing. But with AI helping with coding, automation, debugging, and research, this kind of systematic portfolio diversification is more accessible than ever. The edge is not in making one model perfect. The edge is in building a structure where no single model has to be perfect.

@humblekevin247 Option Omega can be a good SaaS to start with. But I do it now via my own Python scripts and data from Polygon or Databento.

@SystematicPeter How do you backtest premium selling strategies?

@ValueStox Definitely true. So it is important to keep it simple and not to overcomplicate it too much.

@SystematicPeter If you torture the data for long enough, it will confess!

Over 36 years, this very simple Nasdaq + TSX rotation momentum model had only 3 losing years. The rules are almost boring: - 50% capital to US Nasdaq stocks - 50% capital to Canadian TSX stocks - once a month, rank stocks by longer-term momentum - only trade a market when its index is above MA200 That is basically it. The test is survivorship bias-free, fees included, all referenced in USD. Results: CAGR 14.4% max DD -15.62% Sharpe 1.47 Nice example that diversification across simple, easy-to-automate edges can be very powerful. Both markets can be traded mechanically through IBKR in the same trading hours, which makes this very practical for retail systematic traders.

@m_schouten thank you

@SystematicPeter This is a good example of sequencing edge correctly: low-frequency trend + regime filter first, then layering higher-frequency strategies on top. The mistake most traders make is doing the opposite — scaling speed before stability, which turns variance into perceived skill.

Starting traders usually try to go faster. More screens. More trades. More signals. More stress. But speed is not automatically an edge. Very often, it is just a faster way to overtrade, overfit and break your own rules. This green line is my weekly "buy the dip" model on S&P 500 stocks. It trades once per week. The logic is intentionally boring: - Find stocks already in an uptrend - Trade only when the broad market is also in an uptrend - Define the dip on a weekly chart - Check entries on Monday - Manage the trade with SL/PT Just a simple systematic process repeated over and over. I like slower trading because it usually has a better chance of survival. Fewer decisions to mess up. Less noise to react to. Less temptation to optimize the strategy into something fragile. Of course, slow trading has drawbacks. You usually do not want to run this type of model with leverage. It will tie up a lot of capital. Returns take patience. It will not give you dopamine every morning. But as a foundation, it makes a lot of sense. Build the robust, boring portfolio first. Then add faster and more advanced strategies on top. Starting with speed often feels exciting. Starting with robustness is usually smarter. I track this model and its open positions in real time on my blog: Live Trading Models / Buy the Dip (weekly) Use it for inspiration.

@MrBigNatural This one uses my own regime filter, which is very close to C/MA200.

@SystematicPeter Thanks I’ve been trying to implement this. Do you use 200dma/40weekly as your regime filter for stock snd the broad market?