Does Trade Ideas Have Backtesting? The OddsMaker Limit

Yes. Trade Ideas has a backtester called the OddsMaker, it is genuinely useful, and it is available on TI Premium only. The part almost nobody mentions is its history window. Trade Ideas' own OddsMaker user guide states that "typically there are six weeks worth of history in the database." Six weeks. Not six years, not twenty. That single specification decides which questions the OddsMaker can answer and which ones it cannot, and it is the thing to understand before you pay $178 to $254 a month partly to get it.

This is not a criticism of the tool. The OddsMaker was built to answer a specific question well: over recent trading, did the setups my scanner fires on actually make money? For a real-time scanning product that is exactly the right question. The trouble starts when people buy it expecting an answer to a different one.

What the OddsMaker actually does

The OddsMaker is an event-based backtester. It does not simulate a strategy continuously across a price series the way a conventional engine does. It waits for your scan conditions to fire, treats each alert as a candidate entry, applies your exit rules, and reports what happened. You get a win rate, a profit factor, the average winner against the average loser, and a projected return for the configuration you tested.

That design has consequences you should know about, and Trade Ideas documents most of them openly. They are worth reading as specifications rather than complaints.

Verified against Trade Ideas' own OddsMaker documentation on August 19, 2026. Specifications change, so confirm with Trade Ideas before you make a decision on the strength of them.

How far back does the Trade Ideas OddsMaker test?

About six weeks, per Trade Ideas' own user guide, which is roughly 30 trading days. Some third-party write-ups describe a window closer to 40 to 64 trading days, so treat six weeks as the documented floor and two to three months as the practical ceiling. Either way, the OddsMaker measures recent behavior, not historical durability, and no configuration of it will show you how an idea handled 2008, 2020 or 2022.

Put that number next to what you are testing. Thirty trading days is a single market regime. If volatility was compressing over those six weeks, every mean-reversion setup in your scanner will look excellent and every breakout setup will look broken, and both readings will reverse the moment the regime does. That is not the backtester being wrong. It is the sample being short, and a short sample cannot tell you which of the two you are looking at.

Why a six-week sample is the thing to be careful with

Short windows and repeated testing are a bad combination, and the size of the problem has been measured. Bailey, Borwein, Lopez de Prado and Zhu showed in the Notices of the American Mathematical Society in May 2014 that with only five years of daily data, testing more than about 45 independent configurations near-guarantees an in-sample Sharpe ratio of 1.0 with an expected out-of-sample performance of zero. Their point was about backtest overfitting generally: the more variations you try on a fixed sample, the more certain it becomes that your best result is noise you selected rather than an edge you found.

Now shrink five years of daily data to six weeks of scan alerts and consider how the OddsMaker is used in practice. You adjust the entry filter, rerun. Change the stop, rerun. Add a float condition, a price band, a time-of-day window, rerun each time. Twenty variations is an ordinary afternoon's work in this interface, and the tool makes it easy on purpose. The configuration that comes back best across twenty runs on thirty days of data is, on the arithmetic, far more likely to be the luckiest configuration than the best one.

The defense is not to stop using it. It is to decide your rule first, test it once, and treat a good result as a reason to watch the setup rather than proof of an edge. If you find yourself on the eleventh variation, you have stopped testing a hypothesis and started shopping for a number.

Which Trade Ideas plan includes backtesting?

TI Premium only. The OddsMaker is not available on the free plan or on TI Basic, which most reviews call TI Standard. That puts the entry price for any backtesting inside Trade Ideas at $178 per month billed annually, or $254 month-to-month, checked against Trade Ideas' pricing page on August 19, 2026. Holly, the AI signal engine, and the broker-connected automation sit on the same tier.

This matters more than it looks, because the $89 per month figure quoted across most reviews is the annual rate on TI Basic, the tier with no backtester and no AI. Budgeting from that number and expecting the OddsMaker means you are out by exactly double. The Trade Ideas pricing breakdown works through every tier, both billing terms and the add-ons that are listed without published prices.

Can you backtest Holly AI signals?

Not in the way most people mean. Holly is a signal engine whose selections come from Trade Ideas' own pattern matching across a large strategy library, and you do not get to specify, inspect and re-run the rule that produced a given signal. The OddsMaker tests scan configurations that you build. So you can backtest your own scans thoroughly and you cannot pull Holly apart and test the logic underneath her picks.

There is a related detail worth checking before you subscribe for the automation. Holly's own signals cannot be auto-traded. What Brokerage Plus can execute, in a sandbox or live through a connected broker, is your alert window scans. If the plan in your head was to let the AI trade for you, confirm that specific point with Trade Ideas directly.

How do you backtest a strategy in Trade Ideas?

You build the scan first, then hand it to the OddsMaker with entry and exit rules attached. The working sequence is: define the alert conditions in the scanner so the setups you care about fire, open the OddsMaker on that configuration, set your entry timing and your exit rules including the stop and target, choose the window, then read the win rate, profit factor and average winner against average loser in the results.

The step people skip is writing down what result would make them abandon the idea, before they run it. Decide in advance that a profit factor below some level kills the setup, and you have run a test. Decide afterwards, once you have seen five variations, and you have run a search. Once you have twenty saved configurations and a pile of exported results, the comparison has stopped being a backtesting problem and become a data problem, and at that point asking plain-English questions of your own results table will tell you more than another run of the scanner will.

Is the Trade Ideas OddsMaker accurate?

It is accurate about what it measures and it measures a short window with approximated fills. The 1-minute OHLC resolution means intrabar sequence is inferred rather than known, and exits drawn from historical prices without the spread will read better than a live account would have done. Those two effects push results in the same optimistic direction, and they push hardest on exactly the fast, illiquid, high-turnover setups that look most impressive in a short backtest.

So read OddsMaker output as a filter rather than a forecast. It is good at telling you a scan is clearly not working, which is a genuinely valuable thing to learn cheaply. It is much weaker as evidence that a scan will keep working, because six weeks cannot demonstrate durability no matter how good the numbers are. The broader question of what any backtest can and cannot prove is worked through in is backtesting accurate, and the sample-size question specifically in how long you should backtest a trading strategy.

Is Trade Ideas Holly worth it?

It depends on whether you are buying a scanner or an opinion. For a full-time day trader who will be in front of the tape every session, Trade Ideas is a fast, deep, well-built real-time scanner and Premium at $2,136 a year is a professional expense that can be justified by time saved. For someone who trades a few times a month, the subscription is priced against a frequency they do not have, and the OddsMaker's six-week window will not be able to settle the longer-horizon questions they actually care about.

The honest test is the one worth applying to any signal product: would you still open it every morning with the AI switched off? If yes, you are buying a scanner and you should compare it against scanners. If no, you are buying the signals, and signals are the part nobody can make promises about. The Trade Ideas alternative comparison lays out where each approach fits.

What to use when six weeks is not enough

If your question is "does this scan fire on setups that worked recently," the OddsMaker is the right tool and the short window is not really a problem. If your question is "does the idea underneath this rule survive a full market cycle including the bad parts," you need a different instrument, because no amount of configuring will make thirty trading days answer it.

The two instruments answer different questions and using both is a perfectly sensible workflow. Test the durability of the idea across twenty-plus years of adjusted daily data with the trading strategy tester, and if it survives that, use the OddsMaker to check whether your scan fires on it cleanly in current conditions. What does not work is treating a six-week result as the durability evidence, which is the mistake the OddsMaker's own documentation quietly warns you about by publishing the window size in the first place.

Educational analysis only. This is not investment advice, and past performance does not guarantee future results. Pricing and platform specifications were checked against Trade Ideas' published pages on August 19, 2026 and change without notice.