Every speedrun timer shows a projected finish. The arithmetic behind it is as simple as it looks: take your time at a checkpoint, subtract the comparison run's time at the same checkpoint, and add that difference to the comparison run's final time.
projected finish = target finish + (elapsed − target split)
Twelve minutes twenty-two into a run against a split of twelve forty-five, chasing an hour, and you project 59:37. The speedrun pace calculator does exactly this. The number is correct. What it means is the problem.
The assumption hiding inside it
The projection assumes the rest of the run matches the comparison exactly: that you neither gain nor lose another second. That is the only assumption available, since the timer knows nothing about what is coming, but it is not a neutral one.
It treats a lead as a possession. In practice a lead is a probability, and the probability depends entirely on how much run is left. Thirty seconds up with 90 per cent of the run remaining is a weak signal, because 90 per cent of the variance is also still ahead of you. Thirty seconds up with 5 per cent remaining is nearly a result.
This is why the calculator reports the share of the run completed alongside the delta. A delta without that context is close to meaningless, and reading a pace display without it is how runners talk themselves into resetting over noise.
Variance adds, it does not average
If each segment has its own spread, the spread of the whole run is not the average of them. Variances add, so the standard deviation of a total grows roughly with the square root of the number of segments. A long run is proportionally more consistent than a short one, which is why marathon categories have tighter finishing times relative to their length than short ones do.
It also means the back half of a run carries more absolute uncertainty than the front half, simply because there is more of it left. Early deltas are the least informative and the most emotionally loaded, which is an unfortunate combination.
Sum of best is the opposite error
If the pace projection is too optimistic about the past, sum of best is too optimistic about everything. It adds your best ever time for each segment and presents the total as a theoretical personal best.
No such run exists. Each segment best was achieved on a different attempt, often under conditions that traded against another segment: a risky strategy that saved ten seconds here and cost fifteen there, or a lucky drop that will not repeat. Summing independent minima produces a number far below the minimum of the sums, and the gap widens with every extra segment you split on.
So sum of best is a floor that nobody stands on. It is useful as a measure of what each segment is capable of, and misleading as a target.
Reading pace honestly
Use the projection for what it is: a statement about where you are now, expressed in the units you care about. Do not read it as a prediction, and weight it by how much run is left. Split on sections that mean something rather than on every room, because more splits means more noise and more opportunities to misread it.
And expect the distribution of finishing times to be skewed. There is a floor on how fast a run can go and no ceiling on how badly it can fail, so the average completed run is slower than the median, and your personal best is a tail event rather than a typical performance. For the other kind of grind where the tail dominates the average, see the piece on gacha odds.