Risk disclosure

Last updated

What a backtest can and cannot tell you, why simulated results mislead, and the risks of trading that no research tool removes.

Contents

In short

Atlantive is research software. A backtest is a measurement of the past on one series of prices, not a forecast, and nothing it produces is investment advice or an offer to trade. Trading carries risk of loss.

Read this page before you rely on anything the service produces. It is part of the Terms of use.

No advice and no relationship

Atlantive is a tool. It does not know your circumstances, your objectives or your tolerance for loss, and nothing it produces is a recommendation suited to you. We are not a broker, an investment firm or an investment adviser, we are not authorised or registered as one by any financial regulator, and using the service creates no advisory or fiduciary relationship between you and us.

If you need advice, ask a qualified professional who is authorised to give it where you live.

Simulated results have limits

Every performance figure in the service (returns, drawdowns, ratios, win rates, equity curves) is the result of a simulation of trading that did not take place. Simulated results have inherent limitations:

  • they are computed with the benefit of hindsight, on data that was known when the strategy was written or tuned;
  • no money was at risk, so they cannot reflect the decisions, the discipline or the emotions of trading real capital;
  • they depend on assumptions about fills, costs and position sizes that real markets do not honour exactly;
  • a strategy that performed well on the past can perform very differently, and lose money, in the future.

No representation is made that any account will or is likely to achieve profits or losses similar to those shown. The difference between simulated and actual results is often material.

Searching finds luck

Trying many versions of a strategy, or many parameters for one, and keeping the best is a reliable way to find a result that looks good by chance alone. The more variants are tried, the better the best of them looks, whether or not there is anything real behind it. The agent reports how many variants it tried for this reason, and a result should be judged against that number.

A strategy that holds up on data it was not tuned on, at a higher cost, and across different periods and markets is more credible than one that does not; it is still not a promise.

What the simulation simplifies

A backtest models a market; it does not reproduce one. Depending on the data a strategy uses, it may assume fills at prices you could not have obtained, ignore the effect your own orders would have had on the market, understate slippage, spreads and fees, or size positions in whole units that amount to leverage you did not choose. The service states the assumptions a run was measured under with the run; read them before reading its figures.

Historical market data comes from third parties and can contain errors, gaps and days of degraded quality. A result can be wrong because the data under it is wrong.

What a language model can get wrong

The agent is a language model. It can misunderstand what you asked, write code that does something other than what it describes, misreport a figure, state things that are not true with confidence, and give different answers to the same question. Web pages it reads for you are not checked by us and may be wrong.

Review the code a strategy actually runs and the figures a run actually produced, rather than a summary of them.

Futures and leverage

Trading futures and other leveraged instruments carries a substantial risk of loss and is not suitable for everybody. A small price movement can produce a large loss, losses can exceed the money you deposit, and you may be required to add funds at short notice or have positions closed at a loss. Only trade with money you can afford to lose.

Simulated trading challenges

The service can estimate how a backtest would fare against the evaluation rules of proprietary trading firms. Those rules are compiled from public information, change without notice and can be incomplete, and a simulation cannot see everything a real evaluation measures. A simulated pass says nothing about passing a real evaluation, and we are not affiliated with, endorsed by or sponsored by any trading firm.

Exports to other platforms

A strategy exported to another platform, such as a Pine Script export for TradingView, runs under that platform's rules for fills, sessions and data, and can produce different results from the ones measured here. What the other platform does with it is outside our control.

Your decisions are yours

You alone decide whether, what and how to trade, and you bear the results. What we are responsible for, and where that responsibility stops, is set out in the Terms of use.

May we measure this visit?

With your consent, Google Analytics and PostHog count visits to these pages. You can separately allow Google Ads to measure visits from our ads. Visit measurement starts only after you choose. Cookie policy