Testing our own systems in public.
We publish what our testing finds on prop firm challenge rules and Expert Advisor performance, including the findings that argue against the product we sell. Every figure comes from the same source.
TierAlba Research publishes what our own testing finds, including the parts that argue against us. Every figure on these pages comes from the same source: the daily return series of our production engine over 1,221 trading days, built on 244 million real ticks, resampled thousands of times against real challenge rules.
We started publishing this because the numbers advertised in this industry are unfalsifiable. A pass rate with no format attached, no control, and no failure rate next to it describes nothing, and there is no way for a buyer to tell a strong system from a mediocre one run at high risk. These pages are our attempt to be checkable instead of impressive.
The studies
Updated 10 Sep 2026The Zero-Edge Test
We deleted the expectancy from our own Expert Advisor, keeping the volatility and the streaks identical, and ran it through 7,000 simulated challenges. What it passed anyway is the part of every advertised pass rate that belongs to the shape of the game rather than the strategy.
Read the studyRealistic EA Pass Rates
Six risk levels across ten rule formats, with the median time each takes and the failure rate that belongs next to every pass rate. Including the finding that the peak is not at maximum risk: going faster raised failures six-fold and lowered the pass rate.
Read the studyWhich Prop Firms Work With EAs
Nearly every firm permits automated trading, so that tells you nothing. The rule that decides whether a multi-day strategy survives is the drawdown type, and it almost never appears on a comparison table.
Read the studyHow the testing works
Every number on these pages comes from the same pipeline, and it is worth saying plainly what it is and what it is not.
What block resampling means, and why it matters
Shuffling individual days destroys streaks, and streaks are what kill accounts. Resampling in ten-day blocks keeps a bad fortnight looking like a bad fortnight instead of scattering its losses harmlessly across a year. It produces worse-looking numbers than naive shuffling, which is the point.
Why first passage, not final balance
A challenge is not a question of where you end up. It is a question of which barrier you touch first: the profit target or the loss floor. Every simulation walks the path day by day and stops at whichever comes first, which is how the actual rule works.
What we publish that goes against us
A research section that only contains favourable findings is marketing with footnotes. These three came out of the same analysis, and we publish them because leaving them out would make everything else worth less.
The historical sequence was luckier than typical
Observed maximum drawdown against the median across 8,000 resampled orderings of the same trades. The 95th percentile is 49.7R. A drawdown close to twice what we lived through is plausible with nothing broken, and anyone sizing on the historical figure is sizing on a good run.
Faster is worse, past a point
Pass rate from the Standard risk level to the Maximum one. Failures rise from 2.6% to 16.6% over the same step. The fast configurations exist and we could advertise the shorter median time, but we tell people not to use them on funded accounts.
The markets where it did not work
S&P 500, Dow, DAX and FTSE 100, all tested and all discarded. The engine trades the Nasdaq 100 and gold, and nothing else. We also spent weeks looking for a fifth strategy leg, tested a dozen candidates, and rejected every one of them.
Who publishes this
TierAlba builds and operates automated trading systems for prop firm accounts, from Rome, since 2024. The research here is not commissioned or academic: it is the testing we run on our own product before selling it, published in the form it was produced.
That is a conflict of interest and we would rather state it than have it noticed. The defence against it is method: every study explains how to reproduce it on any system, including one that is not ours. The zero-edge test in particular takes about ten minutes on any daily return series, and we would encourage you to run it on whatever you are being sold.