Spire diplomacy/catering tutorial

[MinMax Gamer]

Alright, several people asked me previously about diplomacy/catering in the Spire, and I figured I'd make a short video tutorial with a diplomacy Spire run, and explaining why I make the choices I make. This tutorial is focusing on low-level cities (e.g. chapter 4) as these cannot fight. But negotiation works the same way regardless, just with high level city you may want to switch to fighting at some point.

Hopefully it will be useful for some of you or your friends if they struggle with the Spire a bit

https://minmaxgame.com/elvenar-spire-tutorial-chapter-4-level-1/

https://minmaxgame.com/elvenar-spire-tutorial-chapter-4-level-2/

https://minmaxgame.com/elvenar-spire-tutorial-chapter-4-level-3/

---------------

And apparently by popular demand here is another Spire run, somewhat more advanced. This time this is an end-game city (Chapter 16), so we have demands for all the possible goods. And we will also do some fighting! For diplomatic approach we will use the optimal strategic choices - reflecting costs.

https://minmaxgame.com/elvenar-advanced-spire-tutorial-level-1/

https://minmaxgame.com/elvenar-advanced-spire-tutorial-level-2/

https://minmaxgame.com/elvenar-advanced-spire-tutorial-level-3/

Stay tuned for more videos (check the YouTube channel link in the signature)

 

[PaNonymeB]

Looks like you don't care about avoiding double reds... I usually try to avoid these (because a double red says nothing about the remaining choices) by putting in round 2, yet untested goods (this includes goods who got green, as a green doesn't give any clue about the presence of the good in other positions) for people who first get orange. I also often don't put all choices in round 2, to keep the information "it's somewhere" of one of the oranges for round 3.

 

[Kekune]

I also often don't put all choices in round 2, to keep the information "it's somewhere" of one of the oranges for round 3.

Can you walk me through this one? How/when do you choose to do this, and how does it benefit you? I've always tried to put as many possibilities out as possible, and this contradicts that approach.

 

[ed1960]

If you are dealing with 3 chests and 3 turns you should always be able to clear by going 12312, 23123 patterns or going the 11111,22222 patterns
it when you get to 4 and above chests that duplicates and weeding out unneeded ones matter

 

[PaNonymeB]

The typical (but not ideal) example I use to show my strategy is the following one : 5 options (let's call them ABCDE), in the first round I put all five and get A green, B, C and D orange, and E red.
Then I used to put ABBC in the four remaining places.
Thus I keep the information "there is a D" for round 3, and the greens I'll get will potentially remove some possible positions. With C on last spot I'm avoiding a second red on it, so that possible reds (from either A or B) will appear in positions where I already removed a possibility by the orange in round 1. And, with Bs put in all remaining spots except the last, I'll know if there is a B in the last spot (except if both Bs give green on round 2, but then round 3 is easy with C and D remaining), this way counterbalancing the lack of information in this spot due to a red in round 1.

This is not a ideal case as with recalculating the odds, it turned out that I have with this strategy in that specific setup a 6/23 chance to fail, which is not optimal as CBBA gives 5/23. (In fact, it's like the two Bs transfered the information from orange B in spot 2 in round 1 to spot 5, as "having B" is stronger information than "not having B", so that spot 2 (R1 B, R2 A) lacks of information if there's no A.)
But I compared with "naive" guess ABCD in round 2, and it turned out that with 7/23 chance to fail it's worse than either ABBC or CBBA.

 

[ed1960]

The typical (but not ideal) example I use to show my strategy is the following one : 5 options (let's call them ABCDE), in the first round I put all five and get A green, B, C and D orange, and E red.
Then I used to put ABBC in the four remaining places.
Thus I keep the information "there is a D" for round 3, and the greens I'll get will potentially remove some possible positions. With C on last spot I'm avoiding a second red on it, so that possible reds (from either A or B) will appear in positions where I already removed a possibility by the orange in round 1. And, with Bs put in all remaining spots except the last, I'll know if there is a B in the last spot (except if both Bs give green on round 2, but then round 3 is easy with C and D remaining), this way counterbalancing the lack of information in this spot due to a red in round 1.

This is not a ideal case as with recalculating the odds, it turned out that I have with this strategy in that specific setup a 6/23 chance to fail, which is not optimal as CBBA gives 5/23. (In fact, it's like the two Bs transfered the information from orange B in spot 2 in round 1 to spot 5, as "having B" is stronger information than "not having B", so that spot 2 (R1 B, R2 A) lacks of information if there's no A.)
But I compared with "naive" guess ABCD in round 2, and it turned out that with 7/23 chance to fail it's worse than either ABBC or CBBA.

your idea is fine but how many chests are you working with? we cant discuss strategy without knowing the variables.

 

[ed1960]

IF round one is 5/5 (5 chests and 5 ghosts) and you go ABCDE with the results of "correct, wrong person, wrong person, wrong person, not needed".
You have eliminated "E" so what are you going to do for round 2?

 

[PaNonymeB]

Yes 5/5, that's what I was meaning by 5 options. And then as written I used to do A,B,B,C in the four remaining ghosts. But as said at the end I just calculated that C,B,B,A give slightly better results.

 

[Kekune]

@PaNonymeB thank you for explaining. On higher levels (6+ items) I've withheld a known item in favor of guessing one I had no information about. It never occurred to me to withhold one as you've explained here, though. I'll try that next week.

 

[MinMax Gamer]

This is not a ideal case as with recalculating the odds, it turned out that I have with this strategy in that specific setup a 6/23 chance to fail, which is not optimal as CBBA gives 5/23. (In fact, it's like the two Bs transfered the information from orange B in spot 2 in round 1 to spot 5, as "having B" is stronger information than "not having B", so that spot 2 (R1 B, R2 A) lacks of information if there's no A.)
But I compared with "naive" guess ABCD in round 2, and it turned out that with 7/23 chance to fail it's worse than either ABBC or CBBA.

This is accurate (odds, too! ). But as I mentioned in the video, the tutorial is not meant to demonstrate the absolutely most efficient way of negotiating in terms of total attempts. Rather it incorporates individual variances in the costs of goods, and in this particular case costs in terms of time and complexity (i.e. less thinking -> better ). This makes it less than ideal on pure attempts / unweighted total goods, but I find that this inefficiency is very minor. Think about that - we're dropping success chance from 75% (optimal) down to 65% (naive) - but it is only applicable to a fairly specific group of cases (e.g. 1 hit, 1 miss and 3 redirects). The overall difference is fairly marginal, but you have to keep in mind more different cases (and potentially make mistakes along the way).

It's sort of similar to manual vs auto-fight. You can focus on optimizing your auto-fight setups, but you can always do better by fighting manually - at significant time and complexity costs. Everyone have to weigh whether these extra costs are worth it. I know for me manual fighting is not worth it, but some people swear by it. For the Spire I did write a simulation back in the day when Spire just showed up, and it basically calculated optimal choices at any point in the negotiation. Well, after a little while I abandoned that as it was clunky (I had to enter responses and offers back and forth, sometimes making mistakes along the way) - and it turned out that it really didn't improve my overall results in a material way. YMMV

 

[MinMax Gamer]

...And here is Level 2:

https://minmaxgame.com/elvenar-spire-tutorial-chapter-4-level-2/

 

[PaNonymeB]

This is accurate (odds, too! ). But as I mentioned in the video, the tutorial is not meant to demonstrate the absolutely most efficient way of negotiating in terms of total attempts. Rather it incorporates individual variances in the costs of goods, and in this particular case costs in terms of time and complexity (i.e. less thinking -> better ). This makes it less than ideal on pure attempts / unweighted total goods, but I find that this inefficiency is very minor. Think about that - we're dropping success chance from 75% (optimal) down to 65% (naive) - but it is only applicable to a fairly specific group of cases (e.g. 1 hit, 1 miss and 3 redirects). The overall difference is fairly marginal, but you have to keep in mind more different cases (and potentially make mistakes along the way).

It's sort of similar to manual vs auto-fight. You can focus on optimizing your auto-fight setups, but you can always do better by fighting manually - at significant time and complexity costs. Everyone have to weigh whether these extra costs are worth it. I know for me manual fighting is not worth it, but some people swear by it. For the Spire I did write a simulation back in the day when Spire just showed up, and it basically calculated optimal choices at any point in the negotiation. Well, after a little while I abandoned that as it was clunky (I had to enter responses and offers back and forth, sometimes making mistakes along the way) - and it turned out that it really didn't improve my overall results in a material way. YMMV

Yes, but actually as round 1 is just putting 5 different goods, and round 3 is just putting a possible combination and hoping it's the good one, the only thing to think is round 2, and there are only 20 different setups after round 1. So basically when I see the setup I know what I put in this case. (Even though I can be mistaken in some of these)

 

[Kekune]

the tutorial is not meant to demonstrate the absolutely most efficient way of negotiating in terms of total attempts.

If you run out of other things to do, would you consider making the ultra-efficient tutorial? I think many of us are pretty good at spire, but it's those higher-level concepts we still need to master.

I don't know what I don't know. I think I'm pretty good at spire negotiating, but I learned a new trick today from PaNonymeB. I hope to find more in that vein.

 

[MinMax Gamer]

Yes, but actually as round 1 is just putting 5 different goods, and round 3 is just putting a possible combination and hoping it's the good one, the only thing to think is round 2, and there are only 20 different setups after round 1. So basically when I see the setup I know what I put in this case. (Even though I can be mistaken in some of these)

This gets really cumbersome without supporting tools, and then even with those. The optimal solutions are not intuitive. So in your example you improved odds of failure from 7/23 (ABCD) down to 6/23 (ABBC) and then down to 5/23 (CBBA). But do you realize that this is still a suboptimal solution? IMO the best you can do in this scenario is 4/23, and you can achieve this with ABCC and a few other combinations. So you can see how complicated this can get, and we're talking about just 1 possible setup on turn 2 - and there are many more.

 

[MinMax Gamer]

If you run out of other things to do, would you consider making the ultra-efficient tutorial? I think many of us are pretty good at spire, but it's those higher-level concepts we still need to master.

I don't know what I don't know. I think I'm pretty good at spire negotiating, but I learned a new trick today from PaNonymeB. I hope to find more in that vein.

I don't think that would work out the way you think See my reply above - maximum efficiency is not intuitive. So unless you have a tool that optimizes choices for you, the only way you can achieve it is by memorizing a whole bunch of scenarios, and knowing what is the best choice for each on turn #2. I could probably dust off my Spire solver and showcase a run using the "best" choices, but it is unlikely to be a good tutorial. Because all I'll be able to say is "Well, we see that we've got MATCH, OTHER, OTHER, OTHER, NOTNEEDED", aaand a reference implementation tells us that we should go ABCC on the remaining 4 slots, and that should give us 82% chance of success". I doubt you will learn much from this and won't be able to actually replicate it unless you want to just memorize the scenarios, and it sure as hell won't be entertaining or insightful.

Also, I think you're vastly overestimating "many of us are pretty good at spire". If I had to guess, there are probably grand total of 5 people who would even be interested in something like that

 

[Crow Last Elf]

If you run out of other things to do, would you consider making the ultra-efficient tutorial? I think many of us are pretty good at spire, but it's those higher-level concepts we still need to master.

I don't know what I don't know. I think I'm pretty good at spire negotiating, but I learned a new trick today from PaNonymeB. I hope to find more in that vein.

I don't know if you have checked out @michmarc 's Spire solver or not... it's not a tutorial, but has chances of failure and can be used as a comparison for your own choices.

Here's a post containing a reference to the spreadsheet michmarc's spire solver spreadsheet

@MinMax Gamer -- I think players would want to know how close their own choices are to the theoretically optimal choice. And anything that gives someone an edge in the Spire, with the high cost of negotiating, would be used by serious players, imo.

 

[MinMax Gamer]

...And last but not least - Level 3:

https://minmaxgame.com/elvenar-spire-tutorial-chapter-4-level-3/

 

[Kekune]

I don't know if you have checked out @michmarc 's Spire solver or not... it's not a tutorial, but has chances of failure and can be used as a comparison for your own choices.

Here's a post containing a reference to the spreadsheet michmarc's spire solver spreadsheet

Thank you, I'll see what I can learn from this. You're right that I'd like to see how my choices differ.

@MinMax Gamer I understand I'll never achieve max efficiency without a tool, and that's not my goal. But these tutorials, like others I've seen, seem aimed at people who aren't very comfortable yet with the negotiating process. I think there are a lot of people (me included) who are generally successful with spire and can't gain much from those videos, but who still want to improve. Where I am learning - slowly - is in forum discussions about strategy and odds, some of which have articulated an advanced strategy besides "check the solution tool." I'd hoped to see more of those concepts in your tutorials.

Either way, thanks for making these videos. I'm sure they'll benefit other folks. And now I can read your comments in your actual voice, instead of the one I'd originally made up for you which was not even close.

 

[MinMax Gamer]

@MinMax Gamer -- I think players would want to know how close their own choices are to the theoretically optimal choice. And anything that gives someone an edge in the Spire, with the high cost of negotiating, would be used by serious players, imo.

And apparently by popular demand here is another Spire run, somewhat more advanced. This time this is an end-game city (Chapter 16), so we have demands for all the possible goods. And we will also do some fighting! For diplomatic approach we will use the optimal strategic choices - reflecting costs.

https://minmaxgame.com/elvenar-advanced-spire-tutorial-level-1/

 

[Kekune]

@MinMax Gamer thank you for this! I found the first 3 interesting, and am looking forward to seeing level 2.