Channels

Top 100 Trending New

Categories

Crypto Gaming News Entertainment Music Sports Tech Gambling Bots and mini apps Finance Business Education Lifestyle Health Travel Auto Fashion Jobs Dating Deals and shopping

Stickers

Stickers Emoji Themes Wallpapers

Bots Gifts Tools

Currency

$ USD ₽ RUB € EUR ₴ UAH ₸ KZT ₺ TRY zł PLN ₪ ILS AED AED R$ BRL Rp IDR RM MYR ¥ JPY ₩ KRW CN¥ CNY

Theme

Português (Brasil)

Bahasa Indonesia

Беларуская

Українська

Channels Categories Stickers Bots Gifts Tools

Premium Wiki Glossary Legal

Notifications Settings Log in

NinjaCoding @ NinjaTrader: post #1160 — TG.ME

NinjaCoding @ NinjaTrader@ninjacodingnet · 928 subscribers Open in TelegramOpen

t.me/ninjacodingnet/1160

July 6, 2026 389

More from this channel

Why my strategy research used to end in nothing — and what fixed it For years my process looked like this: build a multi-combinatory system, throw millions of parameter combinations at the optimizer, and hunt for the ones that "work." Sounds like research. It isn't. It's sampling noise. Run enough combinations and pure chance guarantees some look brilliant in-sample. You can't tell the lucky ones from the real ones, so you chase ghosts — they shine in the backtest, crumble out-of-sample, and weeks evaporate. The search space grows exponentially; the information per iteration approaches zero. The fix turned out to be one sentence: An edge hypothesis must be falsifiable. Before I code anything, I now write down: the condition (what setup), the expected behavior (what price should do after it), and — the key part — what test result would prove me wrong. "Breakouts work on momentum instruments" is not a hypothesis. It can't fail, so it can't be researched — only endlessly excused. "After X happens, price continues often enough that entering with a stop at Y is profitable before costs" — that can be killed by data. And a hypothesis that can be killed can also be confirmed. This flips the whole workflow: ❌ Old: search millions of combos → find something that scores → invent an explanation after ✅ New: state one specific claim → test it raw, default settings, fixed basket → it survives or it dies Optimization still exists — but only at the end, walk-forward, tuning an edge that already proved it has signal. Search tunes an edge. Search never finds one. Iterations went from millions to dozens. Every one ends with a written verdict — including the kills. A dead idea with a recorded reason is progress. A folder of "promising" optimizer runs is not. Explain first. Test second. Kill fast. 🧠 I'm now building this whole pipeline — idea → falsifiable hypothesis → screening → walk-forward → SIM, with hard kill-gates at every step — directly into my platform. Will show you how it looks soon. 👀 127·Aug 20 So imagine running a WFO like this, then analysing the results, makeing a few optimization settings tweaks and a few strategy tweaks and then running it again and analysing two different runs against each other. In real life, for human brain its very difficult to make an objective decision what exactly became better and what became worse. Now we have AI to analyse this and talk back to you so and the EOD you get the best possible solution for you live trading. 240·Aug 14

WFO survival) 222·Aug 14 So currently, Claude, GPT, etc. make it possible to write strategies quite effectively. Quite fast, good enough to test and trade. We have definitely taken this step into a new reality. Yes it can go a bit south, but if you iterate over and over, describe whats going on and if you are clear enough with your prompts you will get there sooner or later without touching any code. First, everyone was thinking that this is a game changer, but the real game changer is actually something else. Imagine running even 10+ WFO optimizations trying to make your strategy better and trying to keep intact with whats going on after tweaking it... literally tons of time , and most probably by iteration 2,3,4 you are done with it and it makes you so sick you cant maintain a clear mind as in what did wrong and what did right. Now imagine if an agent was doing this for you? Imagine an agent who remembers and logs every single step you take and analyzes your pathway. This is where the real game changer is in my opinion. 246·Aug 13 [media post] 249·Aug 12

An open web layer over public Telegram — search, stats and channel reading right in your browser, no account, no install.

Not affiliated with Telegram FZ-LLC; all data comes from public sources.

Catalog

Top channels Categories Countries Search

Content

Stickers Emoji Gifts Themes Wallpapers

Help

Tools Wiki Glossary Blockchain Safety Premium Advertisers Advertising

© 2026 TG.ME · G.Media network· Updated 08/23/2026 · v0.2.0 · ba12f6a

Privacy Terms DMCA & complaints Report a violation

Similar channels

Ninjatrader Indicators & Algo1K subscribers

Estrategias Ganadoras de Trading1.7K subscribers

QT finance ®637 subscribers

META_quant470 subscribers

💰 SmiteFX Forex Hub | Free Resources9.9K subscribers

Lock Forex (Free EA & Indicators)16.4K subscribers

Eric Gladyshev - Armaga1.7K subscribers

ninjatrader mzpack bookmap trading soft ...1.3K subscribers