Why we built Paper Trader before we sold it

A trading practice tool that gets the market mechanics wrong is worse than no tool at all — you learn the wrong reflexes. Here's why we ran Paper Trader internally for three months before opening it to outside users, and what that revealed about the gap between "looks like trading" and "is actually trading."

The wrong-reflex problem

The category of "practice trading platforms" is crowded. Most of them are arcade games with a stock ticker on top — you can hit buy, the price moves the way you'd like it to, and the platform tells you how brilliant you are. Then you fund a real account and discover that your fills don't happen at the mid-price, your stops get skipped over in a fast tape, and slippage is real on anything outside the top-50 tickers.

Every habit you built in the arcade-mode tool gets punished by the real market. You learn to size aggressively because slippage was free. You learn to chase because fills were instant. You learn to trust your stops because they never got blown through.

None of those are skills. They're scar-tissue waiting to form.

So when our team scoped Paper Trader, the first non-negotiable was: the market mechanics have to be honest, or we don't ship it.

What "honest mechanics" actually means

Three things, in our scope:

1. Fills happen at the bid/ask, not the mid. If you buy, you cross the spread. If you sell, you give it up. The platform doesn't quietly improve your fills.
2. Stop orders behave like real stop orders. They trigger on a trade through the level, they fill at the next available price, and in a fast-moving tape they slip. We model historic intra-bar tape behaviour where it matters — for a 1-minute candle with a 4% high-low range, your stop is not getting the price you set.
3. Spreads scale with liquidity. A name with $50M daily turnover has a tight spread. A small-cap with $200K daily turnover doesn't. The same $5,000 order doesn't move the first; it absolutely moves the second.

None of this is exotic. Real trading platforms do all of it because they don't have a choice. Practice platforms skip it because the math is harder and the marketing story is worse — "your fills will be worse than the price you see" doesn't sell as well as "trade like a pro."

Three months of dogfooding

Once we had a working build, we didn't ship it. We used it ourselves for three months. Two members of the team trade as a discipline outside work, and a third was learning. We picked names across the liquidity spectrum, ran setups we'd plausibly run in real accounts, and tracked what the platform got right and what it got wrong against what we'd seen happen for real on the same names.

A short list of things that came out of that pass:

• Our initial spread model was too tight on names below $1M daily volume — we widened it after seeing the live tape on three small-cap names.
• Stop slippage wasn't dramatic enough on gap-down opens. We re-tuned the open-print logic based on historic data.
• The journaling layer was an afterthought. By month two, the team member who was learning started insisting it was the most useful part of the platform — auto-tagging entries by setup type so you can review your win-rate per setup, not per ticker. We re-prioritized journaling into the core, not a bolt-on.
• The performance dashboard initially showed the same numbers a real broker shows you — P&L, win rate, average win/loss. That's the wrong dashboard for learning. We rewrote it to surface decisions: how often did you take the setup you said you'd take, how often did you size the way you said you'd size, how often did you bail before stop. The platform now optimizes for the meta-skill, not the score.

None of these came out of a product spec. They came out of using the tool on something we cared about, for long enough to notice what it was failing at.

The journaling layer started as an afterthought and ended up as the most-used surface. The dashboard started as the headline feature and ended up rewritten from scratch. Three months of internal use was the only way we'd have learned either of those things.

What that means for the launch

Paper Trader v1.0.0 is launching. Pricing is still TBD — we want to put it in the hands of beta users at no cost first, get a second wave of corrections in, and then settle the pricing model. The team's strong preference is a flat monthly subscription with no per-trade fees, because we don't want pricing pressure to bias users away from running more reps.

If you're learning the craft and you want a sim that won't train you for a market that doesn't exist, get on the early-access list. We'll open you up to the beta cohort as we have capacity. And if you find something that's modelled wrong, tell us — that's how the platform got this far.