In my naive quest to try to buy coffee from local roasters to support them during the pandemic, I have ended up with lots of coffee that is undrinkable, so I've decided that I'm going to use some of it for experiments.

Yesterday, I burnt through a 250g bag to try to pull multiple shots to investigate what stirring as part of the WDT does. I dialled it in for 15.0g dose shots using WDT, on a profile that was preinfusion until the puck is saturated, then flat 8.6bar pressure. I pulled 12 shots and discarded 6 for not being dialled in, or because I screwed up and hit tare on the scale at the wrong time so I couldn't measure the dose properly.

I used this distribution tool (Levercraft Ultra) and either shoved the prongs right to the bottom of the basket and stirred 30 times, to stir it very thoroughly, or I shoved the prongs about half way down the basket and stirred once, just to distribute the central mound to the edge. Here is the tool:




Here is what the puck looked like after 30 stirs, before tamping:





I tamped with the The Force tamper, flat base, twice. You can debate whether it's a good or a bad tamper, but for present purposes, it's consistent. I also used the Monolith Max grinder with SSP burrs, which are now called "Low Uniformity", I gather. Again, you can debate whether those burrs are good or bad, but, again, it's not the point. I took TDS measurements:




I ended up with 2 sets of 3 shots to compare. Shot 9 was pretty questionable; I can't remember, but I may have screwed up on weighing that, so perhaps that should be discarded. "Total" in the table below is total time, and you can see for shot 9 it was pretty high. Here are the raw results:




None of this tells you what you should do to make espresso, since higher or lower EY might be better or worse for a particular coffee. What this does do is to provide some evidence that WDT techniques do, in fact, make a pretty significant difference. From the above you can see that fiddling a lot, compared with minimally fiddling, results in a drop of extraction yield in the order of 1.5%. Call it 1-2% to allow a margin of error.

A large limitation with the above is that no one WDT stirs 30 times, right to the bottom, with such a device, though it's fast enough that you could if you felt it was beneficial. The reason why I chose to stir 30 times was to try to exaggerate the impact of stirring. Further experiments should be conducted with more sensible things that people are actually likely to do. For now, this is just a result that gives us some vague insight.

We can discuss that, of course, there are a tonne of limitations to this. The sample size is small. I haven't done stats to get a proper margin of error (and the number of results might not even be statistically significant). Perhaps different grinders that clump more will have different results. Maybe changing equipment or coffee will change things. I also screwed up several times in doing this, not taking proper measurements, hitting tare at the wrong time and - the thing that irritated me the most - I set my synching program to push from my laptop to my tablet, so I inadvertently deleted all of the shot history files that I had hoped to extract to graph. I was really disappointed with myself for doing that, because one of the things that I was really hoping to be able to see from putting the six graphs on the one axis was whether one set of shots sped up more than the other throughout the extraction, which is information that we're unlikely to reliable get unless someone graphs it with a DE or an Acaia. Finally, another limitation of course is that I didn't taste anything, since the coffee was gross.

But what I did want to do on this was to move us towards actually sharing and discussing real data. I'm very sick of people making generalisations that are not based on on either actual taste results in the cup or actual data. Discussing actual data is of course difficult, since gathering the data is a thankless chore. Doing the above took me about an hour just to pull the shots, and it's not a particularly impressive data set. But the corollary of my data not being good is not that the opposite is true. It's that we need better data. I'm posting this here in the hope that it might inspire people to give collecting and sharing some data a shot, even if the data isn't great.

Did you happen to notice if, when you "shoved the prongs right to the bottom of the basket" if the face of the basket visually showed more of an uneven pattern of extraction than with only going part way down? I noticed distinct unevenness with my early use of the LeverCraft tool. From Review - LeverCraft Ultra WDT Distribution Tool. I have hypothesized, but not been able to support one way or another than it has to do with the spread of the tines.

I specifically avoided looking at that, because that's too complex. I wanted a simple question, capable of answer, which was just generating the results in that table.

To be very blunt, I don't know how useful any of that information about what beads where actually is, and I'd prefer to pay attention to the results in the cup than something like that, which is difficult to quantify. Is there some study around somewhere, where people have actually correlated any of that with results in the cup? Like I know everyone just loves to assume that "evenness" = better, but that really needs to be put to the test. It might not make a jot of difference.

Luca, thanks for doing this (long overdue) study. I've been meaning to post results on this topic for some time, but it got lost in the shuffle.

So here's a quick & dirty reality check on my system:
DE1+ v1.0 espresso machine, Pavlis water
Baratza Sette 270 grinder, dialed in for ~30s pour time
Black Oak Heartwood blend, 17g dose
EP HQ 14g ridgeless baskets
Decent Tamper v1
fast preinfusion, then 9 bar flat profile at 90C
target weight 34g (brew ratio 1:2)
Atago PAL-COFFEE Refractometer


Procedure:
Grind coffee into two baskets
basket 1: level with finger and gentle shake, tamp
basket 2: brief stir with 4-prong keycap puller WDT tool, brief leveling nutation with tamper, tamp

no WDT on left, WDT on right

Run 1:
No WDT: 26s extraction, Brix 11.13% (21.4C), EY 18.9%
WDT: 31s extraction, Brix 10.55% (22.1C), EY 17.9%

no WDT on left, WDT on right

Run 2:
WDT: 30s extraction, Brix 10.13% (22.1C), EY 17.2%
No WDT: 27s extraction, Brix 10.78% (22.3C), EY 18.3%

no WDT on left, WDT on right

Observations:
* Baskets prepped with WDT take longer to extract to a given brew ratio. I've seen this consistently, ever since I developed this technique. This is probably due to reduced channeling. I do not understand why Luca got shorter pours with WDT preps.

* Baskets prepped with WDT feature significantly increased puck resistance (yellow lines in extraction graphs), for the same reason.

* Baskets prepped with WDT yield visually prettier pours, for the same reason. No WDT on left shows increased spatter.

no WDT on left, WDT on right

* Baskets prepped with WDT give lower refractometer readings (by ~1%EY). This corroborates Luca's results. My WAG: increased channeling on non-WDT baskets causes some overextraction, leading to higher Brix/TDS/EY.

* There were minor taste differences, which (like Luca) I did not pursue. In general, I preferred the WDT-prepped extractions. They tasted less harsh and had more clarity. But this was not blind taste testing, and yes I am biased.

I originally developed WDT for a ghetto-rigged doserless Mazzer Super Jolly grinder back in 2006. That grinder was a clump monster! So my focus back then was breaking up clumps. With my dosered Robur (and Sette), I use WDT primarily for randomizing grinds distribution. Neither of these grinders needs WDT, but I believe both still benefit from it.