So maybe 'quality' is the wrong word.
But let me give an explanation how I see it.
Let's start with a standard search with 1 Thread. Every time we repeat searching the same position to a certain depth, with a certain alpha and beta and with the same aspiration window, the exact same tree is going to be searched. We get the same best move, the same node count, etc. We have a deterministic search. Of course, we all know this.
Here my analogy.
Let's think of a single tree, I choose a cypress tree*, on a meadow. Absolutely no wind, not the slightest breeze! And let's consider it is not growing in the meantime.
This represents our search tree from above. Whenever we look at it, we see the exact same tree.
*Mainly due to
this remarkable post by Chris W.
Now let's come to a SMP search with 4 Threads, YBW type.
Suddenly there is wind blowing over our meadow, and it blows from different directions, and our tree is swaying forth and back, from left to right, etc. Now we always see a different tree, whenever we look! This also helps to visualize, why we don't get a perfect 4x speedup with 4 Threads. Not to speak of 8 and more.
So, every time we start this SMP search, we get different node counts, different best moves and/or at different depths. Non-deterministic. Sometimes we are lucky and the best move is found earlier, sometimes not or it is even found later, compared to our deterministic search from above. These 'sideway movements' swallow maybe 10 - 15% of our search time, and 85 - 90% result in a faster search. Thus the Elo-gain we measure is mainly due to this speedup. You're right that it makes sense to measure the time-to-depth for this kind of SMP search.
Now let's go ahead to the kind of SMP search (Shared Hash Tables (SHT)) Dan and Martin are experimenting with.
Suddenly, we have four cypress trees on our meadow! And at our first look we realize, that it wouldn't make sense to search 4x the exact same tree. So slightly modifying the search depth, aspiration window or maybe even alpha and beta, is a must. Otherwise we would gain almost nothing. Maybe a small speedup, eventually.
So to say, we create our own wind. And eventually we have some control over it by choosing the right modifications to our standard search. As Martin pointed out, there is still room for improvements, most likely.
But of course you see the main difference. We spend more of our time with these 'sideway movements', and thus we cannot get the same speedup like a YBW search. Maybe 50 - 80% for widening the search, and 20 - 50% for a speedup. I can only guess.
Assuming that I'm right, do you still think it makes sense to compare both approaches simply by time-to-depth?
What do you think where the Elo-gains with this SHT search, measured by both, come from? Especially at those lower depths, where a YBW SMP-search does not unfold its full strength!?
But thinking further, it also seems rather logical that we sooner or later get outsearched by the YBW-engine. Which leads me to my idea of combining both approaches.
Why not starting with a SHT implementation, and then switching to a YBW search? Of course, the depth at which we switch would depend on the engine. For a slow searcher, depth 10 or 12 might be ok. Whereas for a fast searcher like Stockfish depth 17 or 18 might be ok. My idea is, that the SHT-search will 'guide' the YBW-search towards the right direction.