A fair difference needs a shared basis and a named reference.
Verdict: A comparison is useful only when every row shares the same basis. The current comparison set covers 13 audited variants, but its totals do not declare a winner.
Check
What it protects
What to do
Shared basis
Symbol, start date, capital, leverage and commission must match
Stop if any field differs
Named reference
Every change is measured against one declared variant
Read the reference before the delta
End date note
Longer and shorter histories are not treated as equal
Do not rank absolute totals when dates differ
Verdict lamps
Risk findings stay attached to each variant
Compare the findings as well as the result
The reference is not an average
Version comparisons use one named build as the reference. Setting comparisons use one named configuration. This keeps every displayed change tied to something a reader can inspect instead of an average that describes no real choice.
The Quantum Titan matrix uses version 3.0 as its reference. The Scalping Robot Pro matrix uses version 4.0. Their notes also say that the end dates differ, so absolute totals are not directly comparable.
An empty cell can be the honest result
Some account survival rows need the same vendor evaluation deposit across every variant. If that value is missing or inconsistent, the comparison omits the row and states the reason. A blank backed by a reason is safer than a precise looking figure built on mismatched assumptions.
The same rule applies when a risk estimate requires fixed position sizing. The Gold Reaper comparison leaves those survival estimates gated because the measured lot policy does not support them.
Zero means something only after the checks
The Quantum Queen X comparison marks its two variants as identical. Every recorded delta is zero and the verdict lamps match. That conclusion is meaningful because the shared basis and required rows passed first.
Read a comparison in that order. Confirm the basis, read any date or omission note, identify the reference, then inspect the deltas and verdicts. The table can show what changed, but the evidence still decides whether the change matters.
Two numbers under every prop firm simulation say what a robot carried into the next day and across a weekend. Lizard shows 0 of 21,524 trades, Gold Atlas 2,188 overnight of which 640 were, on 4 September, counted as spanning weekends. The weekend test then also caught a Monday to Wednesday hold, so a positive count was a ceiling; since 29 September it counts only holds that contain a Saturday or Sunday, and all three profiles we run permit weekend positions so no lamp judges it.
On 29 September 2026 many audits showed drawdowns larger than the deposit they are evaluated on. King TLID, a free audit, reaches 1,263.04 percent of its 1,000 USD vendor deposit, and the same dollar loss shrinks to 12.63 percent of the 100,000 USD test balance.