So here's my data, look forward to seeing what you come up with:
Average PLL execution = 1.29
Solve = 14.07
Cross = 1.33
Cross+1 = 3.91
LL = 4.92
F2L = 9.76
OLL = 2.23
PLL = 2.87
Everything is avg12 except PLL exec, which is average of 21 Mo5s.
Now this one is very interesting, thank you so much for accurately breaking everything down.
I include a jpg of your analysis:
Cross: Excellent. This is the cross of someone who averages 11 seconds. Not a problem.
Cross+1: For such a quick cross, the cross + 1 is a little slower than one might expect. 0.63 seconds could be recovered here by finding the and tracking the first pair during execution of the cross, or even better during inspection time. You have obviously done the work to plan and solve the four cross pieces without much trouble, can you extend this to 6 pieces and not pause before the first pair?
F2L: First pair aside, your F2L is basically OK, perhaps a little slower than it could be, a full F2L of 9.76 seconds is just a little off the pace (9.2), which woudn't be a concern if it weren't for the fact that its supplemented by a lightning fast cross. If your f2l were at the standard of your cross, you would be looking at sub-7 seconds for the full f2l including cross. This would be awesome. See if you can squeeze that extra half second out of cross-f2l transition, and ensure you have a minimum of cube rotations and poor case algs. Personally, I made a chart of every possible f2l case from every possible angle, filled in what I do and then set about reviewing each case to see if there was a better way (anything longer than 8 moves without good reaspn, anything that involved more than 1 rotation, check those that have a single rotation to see if it can be avoided etc etc). Systematic gap plugging like this does add up, its no different to finding a better alg to a PLL you find awkward for example, just a bit more work. I can upload the blank if you like.
Last Layer: Very interesting indeed. Your PLL execution is really fast, a full second quicker than one would expect from a 14.86 cuber. If your PLL execution is 1.29s, I would infer an OLL of around a second, which are certainly heavy duty last layer artillery. Your last layer was in fact a good half second faster than would normally be expected (balancing the half second slower f2l), but given the execution only average quoted, you need to take a look at your recognition/AUFs.
Regarding OLL recognition, if you are keyed into identifying certain sticker to get the exact case, perhaps expand that a bit. For example, where are you looking to tell the difference between OLL 7 and 12 if the lightning bolt has the same orientation and UF is yellow? If you are looking for pairs of yellow stickers, you may have to AUF or move your head/cube, but if you can tell immediately from the which way the FUR yellow sticker is facing, then this is not required. As a practice technique, try getting to OLL and freezing with F, R and U visible. See if there are any cases that come up that require you to look around the cube further, and if so make a note and find other features which telegraph the precise case, so you don't need to look around. It is also worth looking into OLL prediction during last pair insertion. If its a regular R U' R' insertion, the block LU LBU and UB pieces dont change, they simply move anticlockwise around the centre, the pieces at FR and FUR move in behind them (to UBL and UB, thus glancing at the orientation of the yellow sticker on the UFR corner prior to insertion can give info about the forthcoming case, as this ends up at the BLU position, which may need you to look around to see which way the yellow sticker is pointing if you don't see it in advance.) and the pieces at URB and UR remain as they are. For completeness, RFD becomes UFR. Knowing this, you can often predict or partially predict which OLL case is likely to come up, or even if you can only tell the type of OLL that its going to be (say a small lightning bolt for example), you can then make sure your eyes are directed at the right spot to identify precisely which one of the family it is the instant the R U' R' is completed. Anyway even if this doesn't work for you, try to familiarise yourself with all the little details of all the similar cases, so you can identify quickly and confidently, no matter what the orientation of the U face.
If you really want to get hardcore, by knowing precisely which pieces are interchanged by your OLL, you can partially (often completely) anticipate PLL, but few do this. The easier version of this idea is with cases where a group of pieces remain unchanged. For example in the algs I use for cases 28 and 57, the entire L side of the U face remains unchanged throughout. If I notice if its headlights (centre opposite coloured or adjacent?), a continuous bar, a two block closest to me, two block furthest away, or no blocks at all I can partially identify the PLL. For example if its headlight with opposite colour in the centre, I know its a T, H, U or G (d or b) perm, so I am already looking at R side of the last layer when the OLL completes to tell which it is. This is the sort of thing you can get very good at, and I am sure accounts for the minimal recog time of top cubers, even if they don't realise they are doing it!
For PLL, I would look into working on and refining 2 sided PLL recognition which can drastically reduce recog time, and learn some PLLs from different angles so you can reduce AUFs. It seems like you are spending more time identifying and setting up the case than actually executing it! As a practice technique, when you reach PLL freeze looking at U F and R, then try to identify the case from opposite colours etc. It is true that many look very similar, but there are many details you can note (is it opposite colour in the centre of the headlights, or adjacent? adjacent clockwise or anticlockwise? is the block of two next to their opposite colour or an adjacent etc etc). At first this is a slow process, but similar with solving the cross with eyes closed, it really becomes easier over a fairly short time, after all there just aren't that many cases. Or just use the software, which partially removes the temptation to cheat.
Also make sure you know when you begin the alg what the AUF will be in the end. E.g. in a T perm, the colour of the block at FLU and FU will be the colour of the F side of the last layer once the alg is complete. If they are green (say), and the F centre is blue, you can infer that a U2 will be necessary at the end, then tag it on as part of the alg, rather than a separate AUF check.
In summary then, you have some serious technique in play here (cross, alg execution). You need to make sure you are utilising this as effectively as possible by ensuring your recognition skills are as good as your algs. There is some good 2-look PLL identification software that has been released recently, so check this out. Also try and work on that first f2l pair, and eliminate any other inefficiencies in your f2l to take full advantage of your lightning cross skills.
Hope this helps.
P.S. You know when people post sub-1 all PLL videos and someone (usually Escher) comes on and points out that the timing method is a little bit generous and they aren't really all sub-1? I wondered if you are perhaps skewing your PLL times a little bit from what they are in an actual solve with however you are timing them individually in the natural attempt to get the fastest execution times you can. If thats the case, then it would account for why your last layer is slower in practice than your average execution time would indicate, and thus mean that recog is not as much of an issue as the splits suggest. Just a thought.
)
