Papa Soma

@onlybreakouts Damn straight. I've been having an absolute blast with performing forensic reconstruction of various strategies posted on X. Lots of BS out there, but occasionally some gold actually comes along. Full testing regime as you've described applied.

A TradingView equity curve that has never been tested on out-of-sample data is not evidence of edge. It is evidence of curve fitting. This is the core reason most PineScript strategies fail in live trading. Not because PineScript is flawed. Not because TradingView is a bad platform. But because the validation process that separates real edge from backtest illusion is not built natively into the standard workflow. Here is what that validation process actually looks like. In-sample data is the period you use to develop and optimize your strategy. Out-of-sample data is a completely separate period the strategy has never seen during development. The test: does your strategy perform on the out-of-sample data the same way it does on the in-sample data? If yes, that is a meaningful signal. If it falls apart on unseen data, the strategy was fitted to the past. It memorized price history, not market behavior. Walk-forward analysis takes this further. You run the in-sample and out-of-sample split across multiple time blocks: 2016-2017, 2017-2018, up through 2025-2026. A strategy that holds up consistently across all of those blocks is demonstrating robustness, not luck. In professional trading, these are not advanced techniques. They are the minimum standard before deploying anything. The solution for TradingView traders is to use a dedicated layer on top of TradingView that runs both tests natively. Build the foundational breakout model, run the validation automatically, walk-forward it across multiple periods, then export the PineScript and paste it into TradingView. The entire process, from a clean strategy with no filter through full in-sample and out-of-sample validation and walk-forward analysis, can be completed in about 15 minutes for a foundational model. A strategy validated this way gives you a much higher probability of seeing consistent results in live trading. That is the point. Skip the validation. Keep the beautiful backtest. Lose the money in live trading.

@edgeful The blind out-of-sample WFA results verify the edge: Total Trades: 25,329 Win Rate: 66.0% Net PnL: +$7,865,445 Profit Factor: 8.06 Max Portfolio DD: $48.9K Your 3-month sample found the structural pulse. This 3.8-hour vectorized master run proves the current is entirely real.

@edgeful Sizing adapts dynamically to the regime: • Bull (>0.3 signal): 3x contracts with multi-tier scale-outs to catch trend runners. • Sideways: 1x contract steady grinding. • Bear (<-0.3 signal): 0x sizing. Letting the macro pendulum swing over an empty room.

show me a strategy you've backtested that performed better live than it did in the test. we ran one since Jan 11th: ∙ 62% win rate ∙ 53 total trades ∙ 2.931 profit factor the live numbers beat the backtest. here's the full breakdown:

@edgeful The blind out-of-sample WFA results verify the edge: Total Trades: 25,329 Win Rate: 66.0% Net PnL: +$7,865,445 Profit Factor: 8.06 Max Portfolio DD: $48.9K Your 3-month sample found the structural pulse. This 3.8-hour vectorized master run proves the current is entirely real.

@edgeful To optimize computing speed, the code evaluates each signal as an independent round-trip array. If a new signal fires before the previous trade fully walks to completion, the math tracks them separately. An elegant modeling setup that runs millions of rows in minutes.

@edgeful To avoid fighting natural market oscillation, a walk-forward Markov model updates transition probabilities every afternoon. It filters structural noise and coordinates sizing purely based on macro currents, stepping aside entirely during bearish regimes (0x contracts).

@Bober_smart This is basically DMI written in 1978. Wilders had some good stuff

With the help of Claude, I developed an innovative approach to analyzing market volatility and earned $9,670 in 21 days. The model views volatility not as a static indicator, but as a dynamic process of interaction between two forces: > Trend-reinforcing pressure: Pressure reinforcing the current trend. > Mean-reverting pressure: Forces stabilizing the market and returning the price to average values. The mathematical core of the model: V = P(+) - P(-) Key advantages: > Tracking the accumulation of pressure > Risk adaptability > Uniqueness This is not just a theory: it is a working tool that monetizes market complexity into concrete profit. In my article, I revealed the secret and gave access to 10 folders for Claude with the help of which this was created.

@DanielSmidstrup Doctors and Lawyers and Engineers in abundance

USA has ChatGPT USA has Grok USA has Claude USA has Gemini China has DeepSeek China has Qwen China has Kimi China has MiniMax Europe has?

@fabervaaleeng I'm literally doing it right now for myself. Beefy VPS, clickhouse, hermes.

Billion dollar business idea for the builders: Orderflow gives traders an insane data layer: L1, L2, MBO, thousand and thousand of Terabyte of data for thousand of asset class Pattern that a human cannot spot But the market is still missing the real weapon: A software that turns orderflow into quantifiable market microstructure edge. I mean a Claude Code for traders with proper process audit: able to ingest market microstructure data able to test hypotheses fast able to research patterns statistically able to transform discretionary observations into models able to move from data → research → validation → deployment It is an institutional research environment where traders can convert raw orderflow into deployable live market strategies. Data → model → validation → execution. That’s the missing bridge. And whoever builds it properly will own a massive category B2B.

@michaelmalice I would prefer tentacles. All my limbs as tentacles.

Q: Would you rather have tails for legs, or tails for arms? Midwits:

Releasing Early Access to ModMentor today. 30 seats available @ $30/month. Here's what it does. ModMentor is a direct connection between SierraChart and your AI trading advisor. It consists of the Sierrachart bridge, and a standalone application that manages the handoff to LLM.

This analysis was performed at the point of the last advice box. Very nice

On chart plotting of key levels as identified by Mentor, as well as *in this case* key points in the orderflow landscape. Part of the reason I've built this was to short circuit the trading guru notion. Now you have expert advice directly reading you chart and trade metrics.

ModMentor is capable of ingesting orderflow information directly from your Sierrachart chartbooks. Including number bars, and volume profile data, as well as any other studies you may have on your chart.

As well as extracting trade activity data via ACSIL and providing trade stats, it takes all of your trade activity and builds a knowledge base of your trading behaviour. From this knowledge base it delivers insights and actions you can take to improve your performance.

Rather than capturing a screenshot and posting to your AI chat interface, ModMentor captures all chart and study data. Packages the data in a structured payload and sends it off to the model of your choice. Supports Gemini, Chatgpt, Openrouter. More to come.