User:Wyriel

Sailing Advice
Advice:
 * 1) Build Links First
 * 2) Build Normal clears only to
 * 3) cleanup
 * 4) link platforms
 * 5) setup, or extend, a starting trigger
 * 6) Preserve Access
 * 7) Fill, and Drop, Vertically
 * 8) Fill the left or right column of a 2x2, drop into the other one.
 * 9) Horizontally fill the bottom of a 3x2, reducing to 2x2.
 * 10) Don't wait for double-color pieces to fill triangle platforms, unless your 100% positive you have the time/space.
 * 11) Often leave 2 spaces unfilled in a platform instead of just 1.
 * 12) Drop the Wrong Color (and then fix it)
 * 13) Obstacles are worth just as much as Platforms
 * 14) Don't feel bad about getting rid of them
 * 15) Try to combo them if it's easy to do so
 * 16) Avoid having the same color in 4 consecutive columns
 * 17) pay extreme attention to this situation if you have to make it
 * 18) ... or don't finish off the board so you can cleanup should a horizontal clear mess everything up

Meta-advice:
 * 1) Pay attention to grapple
 * 2) Abandon to modify standing faster
 * 3) Also to help with 1).
 * 4) If your doing well
 * ...and your not doing well until well after the first league
 * ...no matter what the indicator or DR says.
 * 1) Stay on station after figurative or literal booches. Same thing as 2.
 * 2) The indicator lies
 * 3) The DR lies
 * 4) Know where the next piece goes
 * 5) Remember what colors/pieces your waiting for
 * 6) If you could place a piece in 1 second, play it slow instead
 * and think about the next piece
 * and what waste needs cleaning up
 * and what pieces are needed to ensure the combo works
 * and where horiz. clears might accidentally occur
 * and then drop it
 * but don't take longer than 10 seconds
 * but don't take longer than 10 seconds

Counter-advice: (contradicts in-game advice or other players' advice)
 * 1) Don't fill up platforms first (Build links first)
 * 2) Don't fill up from the bottom (Build links first)
 * If you can't build a link, then consider filling:
 * 1) Difficult spots (underneath obstacles, poor access...)
 * 2) Link-enabling spots
 * 3) Lower down vs higher up
 * ...because lower down *can* become difficult eventually.
 * However, less priority than links, a), and b).
 * 1) Starting combos with platforms is fine
 * 2) Play slowly
 * 3) no bonus for fast play
 * 4) fast play == less efficient
 * 5) obvious errors like misflips, accidental horiz. clears, etc.
 * 6) subtle errors like 4:42 in Geologist's video MOAR SAILEZ
 * 7) except if grapple is coming up
 * 8) and you can clear the combo by playing fast
 * 9) otherwise just bite the bullet and play 1 mv every 10 sec.
 * 10) or straight-up abandon if you can get away with it
 * 11) Do not break a combo (just) because the indicator starts flagging
 * 12) finish building it the way you intended
 * 13) in general, never panic
 * 14) Simultaneously clearing non-platform balls is worth a small bonus
 * 15) if it's no trouble, clear 5,6,7 instead of 4s (or 8, 9, 10, ...)
 * 16) don't try to separate cleanup clears into independent steps
 * 17) but also don't waste time trying to make simultaneity happen
 * 18) do try to separate normal clears at the beginning, of course

Note that if you follow 5) you will get an excellent on the next duty report in normal pillaging scenarios. This is not an issue standings-wise, and is only a ship performance issue if your the only sailor in the middle of a battle.  In atlantis/flotilla/blockade/etc., following 5) will ensure you maintain top incredible. That is a restatement of the prior sentence: DRs don't lie when they are accumulated over a long period of time (like 10 minutes).

Technical stuff:
 * 1) First-league performance
 * 2) efficient combo in the first 8 moves (source for 8: Shuranthae)
 * 3) drive excess balls on screen down to 0-2, frequently
 * 4) no normal clears: just platforms/obstacles
 * 5) don't even bother with a starting trigger -- takes time
 * 6) unless, of course, it's the only way to get rid of the pieces
 * 7) Do not deliberately add normal clears to the end of a combo
 * 8) when you could get away with not doing it
 * 9) cleanup is a different matter
 * 10) If not board-clearing, don't aggressively cleanup
 * 11) Use excess as starting material for the next combo
 * 12) Abandon with only 1 platform left (and little waste)?
 * I don't do this, but it could be worth it.
 * 1) 1 move every 10 seconds (source for 10: numerous forum posts)
 * 2) idle penalty assessed every 10 seconds
 * 3) probably the penalty is equal to 1 move
 * 4) could also be a constant penalty to points (for a 3 minute window)
 * 5) 3 minute window in 10 second periods (source 3: forum, myself)
 * 6) 30 sessions go into computing standing (source for 30: Jeda and alts in SW)
 * 7) possible to entirely reset standing in about 1-2 hours
 * 8) especially if you can cut board resets out of the equation
 * 9) ... e.g., swabbied ship, blockades, flotilla
 * 10) ... but not usually atlantis, since navvers leave so often
 * 11) Regarding 30, Jeda also comments that it takes only 10 sessions to restore order if
 * the less than perfect SW session was simply very good instead of a total booch.
 * Also, the standing restores gradually, suggesting decaying weights or combining current
 * standing with score immediately (minimizing summation? averaging?)
 * If the combination approach, averaging has issues, less so if rounding is employed.
 * 1) Step multiplier increases without bound (source: myself, c.f. "70 minutes of sailing")
 * (last step of a Vegas^2 is a multiplier of 8 on that step)
 * 1) Approx. weights:
 * 2) normal ball = 1
 * (everything is relative to this; may as well pick 1)
 * 1) extra ball = 2  (5 in a row worth 6)
 * (c.f. "70 minutes of sailing")
 * 1) target ball = 3 (or 4, maybe even 5)
 * (compare "70 minutes of sailing" with "#1 Sailing")
 * 1) obstacle = 3 (or 4, maybe even 5)
 * (Subtle. Junk-clearing obstacles doesn't tank the overall average; see any ult video on youtube.)
 * 1) Efficiency is Points / Moves
 * (regardless of whether or not you use the move)
 * 1) DR is average of Efficiency sampled every 10 seconds
 * (since the last DR, maybe with small overlap)
 * 1) Indicator is the current Efficiency sample
 * 2) Standing is mysterious
 * 3) Child's theory is that:
 * Efficiency is computed once for the whole session
 * 1) Could also be 1 huge DR
 * (average of Efficiency samples over the whole session)
 * 1) Scores are submitted to standing computation
 * 2) Grapple
 * 3) Touching land
 * 4) Leaving blockade/atlantis/flotilla board
 * 5) Abandoning probably carries a small penalty (c.f. bilge)
 * 6) but retaking doesn't forget your Point and Move history
 * 7) which is way better than a cold start
 * 8) Pausing
 * 9) Does not forget your move history for contests: no exploit there
 * 10) Does forget your move, and point, history for the purposes of DR
 * 11) But I think this is only to inform your captain that you were lazing
 * 12) That is, it literally acts like a real pause for the purposes of standing (and contests)
 * 13) Contests/Competitions
 * I, and Jeda, (and others...) have experimented with exploitation strategies that flopped.
 * 1) Best performance seemed to be from completely normal puzzling
 * 2) Jeda's theory: submission is what your session score would be if you abandoned at that DR
 * Which is very different from your DR score, and is dramatically less exploitable.
 * However, super-long sessions of sailing/puzzling tends towards the average, not maximum.
 * 1) Conclusion: Get warmed up on the puzzle and abandon frequently.

Notation
Triple on two platforms beginning with an obstacle: 3n1o4p4p

Vanilla Vegas: 4n 4n 4n 4n 4n 4n

A Whole Board: 3n1o 3n1o 3n1o 3n1o 3n1o 4n4n 4p4p 4n 4p4p 4n 4p4p 6w

A Session: Board1 Board2 Board3 .

weights: 1 1 1 2 (existing) 1 2 2 2 . 1 2 2 3, 1 2 2 4 , 1 2 2 5 X 1 2 3 2. 1 2 3 3 1 2 3 4 1 2 3 5 X 1 2 4 2. 1 2 4 3 1 2 4 4 1 2 4 5 X 1 2 5 2. 1 2 5 3 1 2 5 4 1 2 5 5 X

X - platforms being worth 5 is probably too dominant . - platforms being worth the same as extra balls is very unlikely; also, platforms being worth only double normal balls is somewhat unlikely , - obstacles being worth the same as extra balls is also unlikely albeit, they come as free moves so they get an additional boost anyways to eff.

"reasonable" weights:
 * 1 2 3 3
 * simple (1,2,3)
 * 1 2 3 4
 * platforms as good as extra balls at twice the length
 * 1 2 4 3
 * obstacles are harder than platforms, and get diluted
 * 1 2 4 4
 * simple (1,2,4)
 * 1 2 5 3
 * obstacles are harder than platforms, and get diluted
 * 1 2 5 4
 * platforms as good as extra balls at twice the length, and credit is given to obstacles.
 * 1 2 6 4
 * Too much weight to obstacles?
 * 9, for example, would be totally out of the question, since 3n1o would be equivalent to 4p (it isn't).

Musings
Obstacles are more difficult than platforms, so perhaps their weight is higher, but at the same time, they are "free" points. I.e., 4n and 3n1o are both going to take 2 moves, but not only is the 3n1o worth more points, but also there will be an extra, potentially useful, ball lying around. Or consider 4n vs 4n1o; even if obstacles are only worth 1 the latter is 25% better (4/2 vs 5/2).

Obstacles are a lot more difficult though. One cannot move through them like unfilled platforms, their placement rules are far less restrictive, and indeed they can completely block access. For that matter they can redundantly block very useful corridors, e.g., redundantly block drop paths between platforms. Nigh impossible to incorporate doubled off-color obstacles between platforms, as that requires stacking 6 more balls on top, "even more impossible" to turn this situation into a link between platforms. So, often, one must just junk clear them, and even if they are 25% better than a normal clear, efficiency 1 vs efficiency 1.25 being combined with platforms at step 6 (efficiency 12-24) has pretty much the same effect: tanking the average. From experience, normal clears tank sailing averages much, much, faster than just clearing out all obstacles with a vengeance. Especially so if one attempts to clear out the obstacles using little mini-combos (doubles + triples when possible). Using a higher weight, like 3-5, gives the single obstacle clear about double the efficiency of a single normal clear. If one then lands doubles or triples on these to junk clear them it is clear that the effect on the global average will be greatly mitigated, as compared to the same kinds of combos on normal clears. (consider averaging 62 and 9 counts of 1 vs 9 counts of 2. The first case is 7.1 and the second case is 8.  If obstacles were only 25% better this gap would only be that much smaller.  With the higher weights the gap is big enough that one could easily see it often making the difference between 2 levels of the indicator [if it isn't way beyond sparkly...]).

I have a special place in my heart for big combos, so the most pessimistic set of weights I consider reasonable are (1 2 6 4). Anything higher and I find it hard to believe that one could get incredible/ultimate without breaking platforms. At x4 there is still a huge advantage to combos on platforms, however, clearing away all junk lets one get much bigger/longer combos. Combined with an extra ball bonus of x2, making the combo twice as long starts to compare with platforms, and potentially even beating them because the bound on extra balls isn't as tight as the bound on platforms. (Finite platform spaces gives interesting tradeoffs...). The points at the high ends won't have as good an average as the points on the high end of the platform combos, but the high end can be several times bigger than the lower end, causing the average of the two sides to tilt heavily in favor of the high end.

At a platform multiplier of x2 (with no extra ball bonus), making combos twice as long dominates platforms for the same reason: one can more heavily weight the high end. Even if there is no additional bonus for extra balls, simply putting more of them at the high end than one could place platform pieces has a greater effect on balancing the low end. Keep in mind that the low end includes junk clears used to setup combos, so that the low end can be much larger than it appears; and while the non-platform strategies typically start with junk clearing everything but 1 platform on the board, they don't ever have to make junk clears again. And, for that matter, one could use a hybrid style to platform clear n-1 platforms, and then use the non-platform style for the rest of the session. The platform combos will be limited by the average of junk clears + great combos, because they will keep getting new boards, but in the limit of time the non-platform strategy does not have a significant junk clear component: just a 'great combos' component. In short a x2 platform multiplier just isn't big enough.

For similar reasons obstacles need high weight; but giving both of these weight 3 tends to dominate the non-platform style so heavily that it seems hard not to give an extra ball bonus. As in the beginning of this thread of analysis, with an extra balls bonus of x2 and a platform multiplier larger than that (x3 or x4) there is a kind of balance between the two styles, so that it is possible to be ultimate with the non-platform style. In practice it is hard to build combos truly twice as long as 'pure' platform combos, so a x3 multiplier is feasible, but, a x4 multiplier has a nice symmetry in that building combos twice as long is basically the human limit on the 9 star boards (a pure platform combo on 5 or 6 platforms would be 6-8 steps long; a non-platform combo can somewhat easily reach 9-12 steps, but beyond that gets very difficult without introducing many junk clears (defeating the purpose of getting bigger)).

So, anyways, platforms should have x3 or x4 multiplier, and extra balls x2. As there aren't many obstacles available to combo very late (maybe 5-6) on a board, it is difficult to put bounds on the weights of an obstacle. At x6, using 3 balls to clear it, and doing many simul. breaks, one could get an effective weight on each of those balls very close to x4. As platforms really ought to be the very best things available, it seems reasonable to say that obstacles ought to be less than (maybe even equal to?) x6. I already argued that they should be better than x1; so something like x2-x6 as an outer reasonable range, and x3-x5 as an inner reasonable range. x3 has the nice property that it distributes over extra balls at +1 weight, giving them effectively x3 weight. If platforms are then x4, then one has x1 for normals, x2 for extras or normals + obstacles, x3 for extras + obstacles, and x4 for platforms. This has a nice aesthetic. x4 has the nice property that there would only be 2 important numbers to remember: 2 and 4. Nothing leaps out as 'nice' about x5, except that maybe it gives a bit of credit to the fact that obstacles are harder to deal with than platforms. As stated, x6 has the displeasing property that extras + obstacles would rival platforms, at x4.

Returning to the point, as stated, from my POV the worst case reasonable scenario is weights of (1,2,6,4). Analyzing this weight system applied to lots of different strategies is then an interesting endeavor. In particular, finding big combos that beat platform dense combos is interesting. Cross-checking those results against (1,2,3,3) is worthwhile, and if one really wants to, then one could throw in one more sanity check and compare the conclusion against (1,1,1,2). That is totally unnecessary if the result is that some particular big combo is better than some particular platform dense combo: that will remain true as one decreases the platform and obstacle weights. Reducing the extra ball weight can make a difference, but in the weight system 1,2,6,4 it doesn't tend to be the case that one can beat small platform dense combos using bigger combos unless they also cover all the platforms, and then the point is usually just to increase the multiplier to balance against junk clears used to build the combo. In this case, adding more and more extra balls anywhere in the combo just dilutes the platform density of the combo further without increasing the multiplier chain, and increasing the multiplier chain always increases efficiency (and so dominates adding extra balls). The only reason to consider deliberately adding extra balls would be in some scenario where a great many junk clears were needed and one can't succeed in increasing the multiplier chain; in this case adding these balls at x1 multiplier (instead of x2) is still going to help balance the junk clears, just not as much. The qualitative conclusion will usually remain the same.

Anyways, if some qualitative conclusion concerning build strategies holds in both (1,2,6,4) and in (1,2,3,3) then, in my experience, it holds on the ult list as well. In particular, conclusions I reached based on this kind of analysis informed my deliberate change in style from non-platform vegas^k to full platform small, and earned me #1 from a prior position of #20-#40. At this point I'm investigating a further change in style to full platform big, based on a more in-depth analysis taking junk clears into account.

Alternative Systems
Any alternative scoring system needs to obey the following observations. 3n1o4p4p, if done perfectly all the time, should be a top ten efficiency. It should be possible to beat it by using combos like 4n 4p4p 4n 4p4p 4n 4p4p, even if using a fair amount of junk clears to set it up + leaving a small amount of waste. It should also be the case that the margin of victory -- taking junk clears and waste into account -- should be small, so that a single combo lost to grapple is a great enough loss that the margin of victory is lost, allowing the triple player to remain higher on the standings. Combos like 4n4n4n 4n4n4n 4p 8n8n8n should be competitive with both of the prior 2 (taking junk+waste into account), but even with a perfect record in terms of grapples should lose. Combos lost to grapples need to be a small enough penalty that the player easily remains ultimate, however.

Combos like 4n4n4n 4n4n4n 4n4n4n should be, relatively speaking, pathetic. Some/all of the vegas' text is less than maximum size, indicating an average that doesn't even meet the incredible threshold. Combos like 4n4n4n4n4n 8n8n8n, however, should just barely eke out incredibles/maximum text size. Certainly 4n4n4n4n4n 8n8n8n8n is incredible, but again, ought to be quite inferior to 3n1o4p4p after taking junk/waste/grapples into account. Quite likely one only needs to take into account junk -- this strategy never needs to leave the board, so there is essentially no such thing as waste, but avoiding the occasional normal clear is nigh impossible with combos this large. Combos like 4n4n4n4n4n 4n24n4n4n also happen, and seem to work pretty much the same as 4n4n4n4n4n 8n8n8n8n.