Paste a YouTube link or describe the rules in your own words. We turn them into an exact specification, run them across futures history with real commissions and slippage, and simulate a prop firm evaluation from every possible start date. Then we tell you what happened — including when the answer is no.
The free bot, exactly Passes an Apex evaluation in 30 days or less in our simulation — we write it for nothing. Slower than that and the price still drops the faster it passed. Two prices at every speed: discounted, where we may also trade and resell the strategy, and private, where we will not.Return-to-drawdown 0.31. Our detection floor at this trade count is 0.55. If an edge exists, it is smaller than we can separate from chance.
You never send us a program. You send words, and we build the rules in front of you before anything runs.
A YouTube link, or plain language: entry, filter, stop, target, exit. Any of the 42 supported languages.
We show the exact specification we extracted, plus everything that was ambiguous and what we assumed. You correct it.
Minute bars, real commissions, real slippage. Entry on the next bar open. No exit on the entry bar. Worst case when a bar is ambiguous.
We replay an Apex evaluation from every possible start date — trailing drawdown, daily loss, consistency rule — and report the share that passed.
Extraction is free within your monthly allowance. A check is spent only when you press Run.
Sign in to run a check. A free account includes two checks a month.
YouTube only. It is the one video site with automatic captions we can read, so it is the one we support honestly instead of pretending.
We hold quote files for more symbols than this. This list is shorter, and deliberately so: it contains only the contracts for which we have verified tick size, dollar value and commission. Without those three numbers a backtest produces points, not money — and points are not an answer. The history shown is read from the files themselves, not typed by hand.
| Symbol | From | To | Years | Tick | $ / point | Commission |
|---|---|---|---|---|---|---|
| … | ||||||
Unused checks roll over, up to a cap. Cancel any time.
We write NinjaTrader bots only for strategies that pass a prop evaluation in simulation within a year. The faster it passes, the more the result is worth — and the price moves accordingly. Two prices for every speed: discounted, where we may also trade and resell the strategy; and private, where we will not offer the bot to anyone else.
| Passes in | Implied return / drawdown | Discounted | Private |
|---|
It must pass from every start date we can test, not on average. We replay the evaluation beginning on every trading day in the history we hold. If it gets through on 900 of them and fails on one, it has not qualified. An average of thirty days with a worst case of two hundred is a promise that will not come true for the person who happens to start on the wrong day.
The step is set by the worst case. When we say thirty days, we mean: begin on any day in the whole history and you are through within thirty. Not usually. Any.
The only start dates we leave out are the ones we cannot judge. A run that needs thirty days cannot be scored against the last two weeks of data — the window does not fit. Those dates are excluded as unmeasurable, never counted as passes.
And here is what it still does not mean. All of this proves the strategy held across the whole history we have. It does not prove it will hold tomorrow, and we do not claim it will. What we refuse to do is build a bot on something that worked once, years ago, and has not worked since.
Discounted pricing grants us the right to trade the strategy and to offer a bot based on it to other customers. Private pricing does not. On both, we retain the right to trade it on our own account — stated here rather than hidden in a policy page. Every bot is delivered as is, for use at your own risk.
Everyone selling backtests omits this page. It is the reason to trust the rest.
It describes one path the market already took. Ours is measured against a control: 100 random entries with the same trade count and holding time. A result that cannot beat coin flips is reported as exactly that.
We injected known edges into synthetic strategies to find out where our filter goes blind. It detects roughly 0.55 return-to-drawdown out of sample and misses 0.30. So we write "no edge could be detected", never "this does not work" — the difference is not politeness, it is accuracy.
Inside one minute we cannot know whether the high or the low came first. We run both orderings. When they disagree, the report says inconclusive instead of picking the flattering one.
Prop firms change their rules without notice. We date every rule set we used and print the date in the report. Verify with the firm before paying an evaluation fee.
No, and we would rather you did not. We build rules from a bounded list of things a strategy can do, and we show you the result before anything runs. We never execute code you send us — running a stranger's program on our machine is a hole, not a feature.
Automatic captions. YouTube produces them for practically every video; other platforms do not. We support what we can actually read rather than advertising what we cannot.
Only what you have the right to share. Most marketplaces forbid both decompiling a product and handing it to a third party for analysis. Describe how it trades instead — that is what we test anyway.
You get the full report and the check is spent. A negative answer is the product. We would rather tell you before an evaluation fee does.
On the discounted tier, yes, and we may sell a bot based on it. On the private tier we will not offer it to anyone else, but we still keep the right to trade it ourselves. We say so plainly because a promise we do not intend to keep is worse than an uncomfortable sentence.
This answers questions about how Karivox works, what the plans include and what the documents say. It cannot advise on trading, evaluate your strategy or predict anything — and it will say so rather than guess.
Karivox