So it seems. But I don't know how to fix that.

It's because you embedded the sgf content directly (which is quite large with all your comments, ~10KB), and the eidogo download link tries to put all the sgf in a http GET url and exceeds server limits (often 8 KB). If you hosted the sgf elsewhere and linked the url, or uploaded it as a file to L19, then I expect the download would work.

It's the end of the 2017-2018 table tennis season, which frees up at least my Wednesdays, which happens to be the club night organized by my new friend Ian Butler. His enthusiasm has been contageous so I thought I'd pick up the study rhythm again, for as long as it will last.

I'm following my own study program, which tells me to play 10 blitz games and not resign. Whatever happens, play it out till the end, aspiring to come back when behind and maintain the lead when ahead.

See https://senseis.xmp.net/?DieterVerhofst ... ingProgram

Given that such an attitude may not be welcomed by human players, I may play bots only. Here's game 1:

2nd blitz game of theme 1: develop fighting spirit by not resigning (at least not too soon)

Somehow I seem able to confirm a 3d rank on KGS, which I was never able to do before. So it must be an effect of the small sample of games I've played over the past months, or KGS has become a bit weaker. Chances that I have become stronger by mostly doing L19 reviews are slim.

There's a bit of blitzbluff to be enjoyed at the top, reviving a group in seki. Otherwise, fairly happy with the level here.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

My two main guides in life:
My mother and my wife.

Everything with love. Stay safe.

Awesome game and congratulations on the 3D status. That's only good for our Go-club when you drop by, adds to our prestige

I'm happy I played a little part in your playing more often again! That seki is brilliant, by the way!

Just curious... The second game, at 0:00 + 5x0:10, most people would say is blitz. The first, at 10 + 5x0:30, do you also consider that blitz? I remember a great page you had on SL some time ago where you preferred bonus time. (I do too, but have few chances to play that way.)

No, the settings weren't blitz. But I played it like blitz, not using my thinking time.
Well spotted though!

Series 1: fighting spirit, is done.

I'm not too happy overall: my fighting spirit looks more like desperado blitzing, which is not what I intended. The last game is symbolic: I was 60 points ahead and still lost through careless play. Not a display of Dan strength ...

Focus is thus on steady play. Series 2 will be about time management: play 10 games at 5/30, use my time and the opponent's. Play steady moves.

While I'm playing more games, now with the focus on time management, I also picked up an old idea: "being Otake Hideo".

I chose Otake as my role model because he's known for playing shape and using thickness in the endgame to nibble away the opponent's points. His style may be easier to emulate than the sharpness of Lee Sedol or the imagination of Go Seigen, although I'm probably underestimating Otake's sharpness and imagination.

Anyway, here's the idea: I pick a game by Otake and for every move choose 3 alternatives.

My score of "being Otake Hideo":

1. ratio of his move being among my 3 chosen alternatives: 62/94 ~ 66%
2. ratio of his move being my first choice: 37/94 ~ 39%

Incidentally, this was not his best game: he gave up in the endgame. So this was more like "being behind with Otake Hideo".

Oh and don't look at 1,2,3 but at A,B,C. I did a similar exercise with this game a long time ago.

Game two in "being Otake Hideo":

top 3 hit rate: 57/93 = 61%
top 1 hit rate: 40/93 = 43%

Knotwilg's pro games endgame study

I'll post some examples later. First I'll explain what I'm trying to do

Idea: use close professional games for endgame study
Procedure:
- browse the game until I find
a) all important groups to be stable
b) no major areas left to develop
- then make the following evaluation
- who has sente (say: Black)
- which (major) sente moves does Black have
1) threatening a group
2) threatening to enter a territory
3) NOT ~1 point sente moves, which are really ko threats
- which big gote moves are there
- and so what are the moves I can expect (being: Black's sente and the biggest gote)
- repeat this evaluation after most of the expected moves have been played, or the game has taken an unexpected turn for me

Lessons learnt:

At my level there is really only one major lesson to be learnt and that's "Endgame = Life & death". Wherever my analysis deviates most from the pros', it's because I wrongly evaluated a group to be so alive that there were no immediate sente moves against it.

A variation on the theme is connectivity. Pros are very clever in finding big moves that affect the connectivity of a dragon. Both are examples of how pros squeeze out the available major sente before turning to a big gote. Every game I've analyzed has one or more big gote that are played much later than I had expected.

At a higher level, there are probably many more lessons to learn, such as fighting spirit, precise evaluation, kos (which can severely disturb the equilibrium) and other potential swaps.

The amateur's focus on "big gote" tells me that my/our theoretical framework overly emphasizes techniques to count the value of moves at the expense of techniques to find sente. It's sufficiently clear that we should play big sente first but we are not really offered a framework for being relentless at finding them.

Good point.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

My two main guides in life:
My mother and my wife.

Everything with love. Stay safe.

Fueled by the availability of LZ, I looked at some older studies of mine. One of my study approaches is "Being Otake Hideo" after the Malkovich movie. I replay a game by Otake and pretend I'm him. I pick 3 choices, then see if Otake played one of those. Later I tried to replicate this in games: "What would Otake do?"

I chose Otake because he presumably has a style (there, I said it) which is easier to emulate by amateurs, since he goes for "good shape" more often than other pros, or so I heard. So, as a matter of substituting the hard reading with narrow margin for error, with the "lazier" shape approach, I pick Otake-san as my role model, probably entirely unfairly and apocryphally so.

Now I enhance this study with LZ's top 3 anytime I choose a move triplet.

Some differences I observe between Otake and LZ:

- LZ uses similar probes as Otake, but Otake then immediately plays elsewhere, while LZ tends to settle the shape further before playing elsewhere
- LZ is not very fond of a capping play to attack a group, blocking its way forward, with merely new influence as a result
- LZ does not let the opponent off the hook, while Otake often injects what looks like an endgame play; in that sense Otake sometimes plays more AI-like than LZ, taking the points elsewhere instead of continuing the attack

Like before, I have observed heavy swings in LZ's appreciation of a pro game involving Otake. It can favor White up to 70% then dislike a couple of moves so much that the situation is reversed very soon.

I learned that play from Sakata, and put it to great use in my games. But, many years later in studying Okigo Jizai I noticed that Hattori seldom used it, generally preferring a sideway attack. Was that Hattori, the times, or handicap strategy? I couldn't tell. But now I have noticed that the top bots like sideways attacks, as well.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

My two main guides in life:
My mother and my wife.

Everything with love. Stay safe.

Things are converging.

I analyze games with LZ, seeing variations through her eyes.
Instead of guessing moves in pro games, I try seeing variations myself. Then compare them with LZ's.
Next thing is playing a game and trying to see variations, instead of seeing just moves, or talk myself into something.
After 24 years of playing Go, will I be able to acquire the real language it speaks?

Handtalk.

(PS: I will possibly refrain from posting the results here, since that only detracts from speaking the real language).

The first game where I consciously applied "seeing sequences" and next reviewed it with LZ was instructive in so far that LZ saw sequences I didn't (of course) and at least one of them I still remember: a tenuki variation of a joseki. My opponent, a lesser bot, also saw sequences that I didn't, but then in terms of a resisting variation in a tactical fight I had not considered

The biggest surprise is that LZ saw me falling behind steadily, up until a rate of 25%, while I thought I was consistenly ahead, harassing 3 groups in my sphere of influence. LZ thought those groups could handle themselves well, my surrounding stones had weaknesses and the territory I had given in exchange exceeded the potential of the attacks by far. Eventually I won in a big way, killing a group. However, that was more due to an oversight by the bot than a necessary consequence of the splitting attacks.

The biggest upside is that "seeing sequences" does slow me down in automatic play and makes me evaluate moves in a more disciplined way.

I'm going to make "seeing sequences" the new headline of my study journal. This change of perspective has been mostly induced by the advent of Lizzie as an analysis tool. Lizzie will become a centre piece in my analysis of my own games, people reaching out to me for game reviews and when reviewing pro games.

The idea is to use AI language in reviews. AI language is void of terms. It speaks in candidates, winning probability and sequences. The human player also considers candidates, sees sequences but doesn't evaluate the result in terms of probabilities. Rather, we perform positional judgment, taking into account territory, strength, influence, life & death, sente, aji ... Incidentally, both AI and (strong) humans value efficiency of stones, for which we seem to have a common intuition. "Seeing sequences" is the common technique, which I want to exploit more.

My own games
1. Play a game
2. Review it and record the sequences I saw plus evaluated
3. Review it again with Lizzie and add sequences it saw plus its probabilities
4. Learn which sequences I should start seeing, which to stop seeing and which to keep seeing

Pro games
1. Review a pro game
2. See sequences and evaluate them
3. Interpret the pro's choice
4. Review it with Lizzie: alternatives, probabilities, sequences

Other people's games
1. Review them with Lizzie, add most relevant sequences (I won't add my own anymore)
2. Continue to find root causes for their mistakes and recommend focus areas

"Seeing sequences" in study typically happens when studying joseki, so this is a topic which I have consciously neglected in the past but will pick up again. It also happens in life & death problems and endgame problems, so this is something to continue. It is harder in specific middle game situations, where intuitive judgment prevailed and where I see the biggest difference with Lizzie's evaluations. I want to study some key books like Attack & Defence with Lizzie and learn new ways of evaluating middle game positions.

So here we go.

I'm guessing from the fact that you chose a new label ("Seeing sequences") that what you are doing is different than what people usually refer to as "reading". Could you clarify the difference? I don't see it from what you've written.

There's no real difference in what happens: it's a fancy way of naming it, but I also use it in opposition to the language we use to evaluate positions. Up until now, I considered "reading" a means to an end, whereas now, I consider "seeing sequences" as some kind of end. Of course we still apply an evaluation at the end of the sequence, but perhaps there is something in the sequence itself, "seeing a good sequence" vs "seeing a bad sequence".

It relates to the discussion about high level terminology and expert understanding, where Sorin German argued that the high level language that pros use to communicate with each other are the sequences themselves. They can probably convince each other of the superiority of a move by showing a sequence alone, rather than debating the relative merits of the end result.

Otherwise, it's indeed about reading. I want to do more active reading in games and compare that reading with the reading of experts. Nowadays, AI tools like Lizzie give full time availability to experts "seeing sequences" that you could have seen. They don't tell you whether the thickness is superior.