Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
tmorrow

A quantitative answer to "How many attempts on a map before giving up"?

Recommended Posts

 

A somewhat odd topic, I know but it's without precedent, having coming up before in various guises on these forums. I do believe that more than one keyboard and mouse has borne the brunt of a bad doom session that has gone on too long.

I used to beat my head against the wall of insanely difficult doom maps until I broke my keyboard over my head, chewed up my mouse and rage quit. You might say I had an anger/frustration management issue! Of course the root cause didn't lie in doom or the map, it was all about my own personal issues that needed to and were explored and will be revisited from time to time.

Ok, so you love doom, you like to play maps, and you like a challenge. So I pose the question "How attempts on a map before giving up?"

One good qualitative answer is "While you still enjoy it and before you get angry and frustrated". But that answer doesn't address the disappointment that nags you when you didn't complete your goal and the regret of wasting so much time on a failure (this is why you want to celebrate the hell out of your successes, humans are masters in the art of failure).

At some point a while back, with mouse wires dangling and frothing at the mouth, I stepped back, settled down, pondered the situation and developed a simple metric to help me estimate how successful I might be on a difficult map based on simple inputs gathered up front.

I'm probably not the first and others may have better ways to approach the problem (that I'm eager to read about), but I'd like to share with you a numerical approach to this problem that I've been using with some success.

Let's assume you've played the map enough to know what it involves and have identified specific difficult sections and you know how enjoyable the map is to play overall.

As rough measure decide on N, the number of times you think you're prepared to attempt the map in faith with a happy frame of mind. You can iterate on N as you play the map, changing it up or down with new insights. At some point you'll settle on a number. You'll get better at doing this over time. Also baked into N is your understanding that in doom, rng is not always on your side and sometimes shit happens, you are happy to roll with those punches to a point.

Now break down your map to n, isolated difficult areas where you usually might fail and assign a probability to complete each of the n areas, p_i for the i'th one based on practice sessions where you repeatedly reload and retry each of the difficult areas. This is part of your up front investment in the map. If you've done your analysis well, you will have identified small segments that you can quickly retry without taking up a lot of time. You can iterate on the p_i, your estimate will get better. You understand that you can fail outside of these areas, but a lot more rarely then within the trouble areas.

Then from basic probability theory, take the product of your probabilities p_1.p_2...p_i...p_n = P to work out the probability of success assuming independence of events and that your practice session estimates are reflective of your in game performance (not always but is usually the case).

 

Finally we can work out a number of successes estimate (NOSE) or the expectation by assuming a binomial distribution. This requires independence of each trial and isn't quite true since you will probably get a bit better with repeated attempts at your map, but we can work with it. It will produce a lower bound for sure so you can replace the = with a >= if you like in the following estimate:
 

If NOSE is the number of successes estimate, then NOSE = N.P = N.p_1.p_2...p_i...p_n.


By way of example I'll apply the NOSE to some difficult Super Mayhem 17 maps I've been attempting uvmax's for, playing with this month's DWMC.


map11: Outrageously hard map. Five difficult areas identified with probabilities 1/4 (2 archviles), 1/3 (revenant trap), 1/3 (cybie fight), 1/4 (lava room), 1/10 (final room). Happy to try 50 times. Apply NOSE = 50.(1/4)(1/3)(1/3)(1/4)(1/10) = 50/1440 = 0.034 = Forget it, map too hard. With 1440 attempts for a likely exit, it's out of my league.

map22: Hard map. Four difficult areas identified with probabilities 1/3 (icy building), 2/3 (ice rink), 2/3 (archvile, hellknight pond), 1/3 (water channel). Happy to try 50 times. Apply NOSE = 50.(1/3)(2/3)(2/3)(1/3) = 200/81 = 2.5. First uvmax on 23rd try, good result.

map26: Tough survival map. Four difficult areas identified with probabilities  1/2 (bfg room), 1/2 (east wing), 1/4 (pillar room), 1/3 (final room). Happy to try 50 times. Apply NOSE = 50.(1/2)(1/2)(1/4)(1/3) = 50/48 = 1.04. So I could roughly expect 1 success in 50 tries, not great but I was happy enough to go for it. Sadly after the 50 attempts I got no successes although I did get 2 almost's, failing twice with less than 20 enemies remaining. I knew going in though that around 1 success was what I could expect.

My experience shows that NOSE is far from perfect but it is a good, upfront ballpark estimate. I've been using NOSE for quite a while now with excellent results. My expectations are more realistic, I waste less time on maps that are too difficult and there's a lot less anger, frustration and disappointment. Your mileage may vary but I humbly offer it up to the wider community in the hope that it will be helpful to others. Even better, may it save keyboards and mice throughout the doom community.

Edited by tmorrow : stated the implicit binomial distribution assumption

Share this post


Link to post
2 minutes ago, Linguica said:

Sounds like you applied the Drake equation to Doom.

More like a micro, mini lemma to Drake's equation but you're right, the same principles are at play.

Share this post


Link to post

I like to apply the question of "am I still having fun?" If I'm not... why bother? I like a good challenge, but when I get no where or its just too overwhelming, there are other things that will be more enjoyable.

 

When reviewing a map I'm incline to give a map one star that I didn't finish. Not only because of my own failed attempts, but a map layout so confusing I couldn't figure out what to do or where to go next. Even if I don't finish a map due to its difficulty, I can still see the time and effort that goes into such things.

 

I'm reminded of @fonze 's map having to brawl with lost souls for 20 minutes to complete it. I still saw it as a giant challenge, but still manageable because I knew what I had to do, I just had to get better at doing it. Meanwhile, with megawads like Eternal Slumber Party, I circle strafe around 200 barons, 100 cyber demons, 100 spider masterminds 50 arch viles... then I look to see oh... 3,000 enemies to go on this map alone and then I say... nope. Done. I have wasted 20 - 40 minutes on this map alone. Time to move on. While the numbers are against you, its not even difficult, its just not fun.

Edited by geo

Share this post


Link to post

UV out of my league? Try HMP. HMP still not good enough? HNTR if I have to. I have to be sold on the aesthetic, the architecture, the atmosphere, the layout, whichever, whatever helps it stick out. But most importantly, if I'm just simply not having any fun whatsoever, then it's time for it to be sent to the trash.

Share this post


Link to post
1 hour ago, cyan0s1s said:

UV out of my league? Try HMP. HMP still not good enough? HNTR if I have to.

That isn't an option with a lot of wads that don't provide for them. The ITYTD always works though with double ammo and taking half damage and is a good option to keep in mind if NOSE comes up with a low probability for success.

 

My main point is you can do decent job quantifying your likelihood of finishing any given map on whatever chosen difficulty if you break it down the way I'm suggesting. You get a decent estimate based on your map analysis and then you can decide whether you want to persist at the difficulty or perhaps drop to a more suitable difficulty so you don't get frustrated. The method works whether applied to speed runs where you are interested in time splits, map completions or whatever your actual goal is.

Edited by tmorrow : corrected the difficulty level

Share this post


Link to post
41 minutes ago, geo said:

Meanwhile, with megawads like Eternal Slumber Party, I wade through 200 barons, 100 cyber demons, 100 spider masterminds 50 arch viles... then I look to see oh... 3,000 enemies to go on this map alone and then I say... nope. Done. I have wasted 20 - 40 minutes on this map alone. Time to move on.

 

In this case you don't even need NOSE, extrapolation on a subset works fine.  You took a good sample of the enemies on the map, 450 or so that took 30 minutes or 4 seconds per enemy and extrapolated that to the entire map so (3000*4)/(60*60) = 3.33 hours of play, far too long for you.

 

NOSE is used more for when your goal is to finish a map at some difficulty or at some speed and you can divide your goal into subgoals that you can assign a reasonable probability to. The idea is that it quantifies how likely your goal is or how many attempts you can expect to need to fulfill it.

Share this post


Link to post
4 hours ago, tmorrow said:

NOSE = N.P = N.p_1.p_2...p_i...p_n.

Woah!

 

Here's my formula:

1. Try.

2. Go to Step 1, until I start to get bored/frustrated.

3. Apply the most minimal cheat, as needed to get past the frustrating part.

4. Go to Step 1.

 

The whole reason I'm playing is to have fun. I can live with giving it a good try or two, then using the occasional cheat to let me get right back to having fun.

 

Now, if I was intent on beefing up my skills, I might increase my try count a bit. But I don't smash up my equipment...cause it doesn't care.

Share this post


Link to post

For the original question, it very much depends on what exactly am I playing.

 

If it's a map I know is insanely difficult beforehand then I don't even bother playing it just to destroy my synapses. If I am taken by surprise by a level down the road that's part of a wad I want to finish well, I take my time and sometimes spend a few hours save scumming until it's finally done. I never cheat or change the difficulty, so if I start with UV, it's going to stay UV for the entirety of the wad. And if I see I have just absolutely no chance of winning, welp, it's time to give up, if it takes more than 40 tries and way over and hour with little to no progress that's an indication I have no other option but to give up and move on to something else.

Share this post


Link to post
9 hours ago, kb1 said:

Here's my formula:

1. Try.

2. Go to Step 1, until I start to get bored/frustrated.

3. Apply the most minimal cheat, as needed to get past the frustrating part.

4. Go to Step 1.

 

Excellent algorithm. The only thing it doesn't cater for is an up front estimation of likelihood of success and possible disappointment/regret. You can add step 0 to address the disappointment though:

 

0. Go in with no expectation that you'll complete your goal. If you end up succeeding it's a bonus.

 

Share this post


Link to post
6 hours ago, Agent6 said:

if it takes more than 40 tries and way over and hour with little to no progress that's an indication I have no other option but to give up and move on to something else.

 

Ok you have N=40 but you don't have an up front assessment of likelihood of success. Having that can give you an early out to some maps you might mistakenly assume you could pull off in 40 tries.  Some people don't realise that a map with say 3 hard parts, 2 with a 50-50 chance and one with a 1/10 chance to get through means around 40 attempts are needed to expect a single success on average.

Share this post


Link to post
4 hours ago, NinjaLiquidator said:

Till it breaks

 

That is the most brutally succinct algorithm I've ever heard! Now what is breaking? Your head on your keyboard or the map finally succumbing to a lucky run?

Share this post


Link to post

Personally, if I were trying for a casual max, and I forecasted odds as low as the ones you listed for map11, I'd set up saves, work out better strategies, and practice more. It would take less time to eventually get an exit in the end anyway. Even if I ultimately don't run the map, I'd be gaining skills and strategic knowhow that can be applied elsewhere. Besides, I find that sort of practice intrinsically fun and satisfying. My personal experience even in pretty hard maps is that odds of 10% are often a sign of less-than-ideal strategies or less-than-enough practice, either or both. Rather, I would expect those numbers for stretches of a dicerolling speedrun. 


Additionally, attempt count is an unreliable measure, because not all failed attempts are equal in terms of time spent. Compare a ten-minute map with possible deaths at 1:00, 3:00, 5:00, 7:00, 9:00 (abstracting for simplicity) corresponding to fights with (9/10)(9/10)(9/10)(9/10)(1/10) survival likelihoods to one with the same parameters and (1/10)(9/10)(9/10)(9/10)(9/10). In the first case, an average attempt takes 7.25 minutes, and in the second, it is 1.75 minutes, despite the same overall odds of success*. 

 

 

*For such reasons, it's also sensible to reset early upon non-death outcomes unfavorable enough to make the rest of a run too improbable to be worth it. Early resets are substantially cheaper than late ones in terms of time spent. Resetting 15 seconds deep compared to 300 isn't just 20 times cheaper, but rather a proportion reflective of the real-life time it would take to get 300 seconds into a run desirably (perhaps many times more than 300 seconds) compared to 15 seconds (probably 20 seconds or so, some cushion added to account for prBoom+ load times). 

Edited by rdwpa

Share this post


Link to post
15 hours ago, geo said:

I'm reminded of @fonze 's map having to brawl with lost souls for 20 minutes to complete it. I still saw it as a giant challenge, but still manageable because I knew what I had to do, I just had to get better at doing it.

 

I remember you reviewing this one. ITYTD and HNTR are the toughest two difficulties on that map. HMP and UV have more 'typical' monster compositions, and HNTR is the alternative 'lost soul and PE spam' mode. 

Share this post


Link to post
1 hour ago, rdwpa said:

Personally, if I were trying for a casual max, and I forecasted odds as low as the ones you listed for map11, I'd set up saves, work out better strategies, and practice more

 

An important observation and setting up saves and working out better strategies is exactly what I do. If I can't get those probabilities down for each segment to below 1/5 or so it's probably as sign that the map is beyond me. It's part of what I call the up front investment which helps get me determining the segment probabilities. Success s times after p plays of a section means a probability of s/p. You want p large enough to have exhausted most of your good ideas by then. There is a little more baked into my probability determination than I explicitly laid out. Thanks for bringing it up.

 

1 hour ago, rdwpa said:

Additionally, attempt count is an unreliable measure, because not all failed attempts are equal in terms of time spent.

 

A good point and I agree it's not a great measure of time spent but it's worked well enough for me up to now. Sure, I could get a better up front estimate of time spent by calculating the chance of getting to the i'th hard section and adding a requirement to estimate time for each section. That's quite feasible but does complicate the formula a bit. I haven't felt a need to go that far yet as the the rough rule of thumb NOSE measure has been working for me quite well so far.

 

Again the point is not that NOSE is perfect, it's far from it but it is a  simple quantity well worth calculating to help decide if you are ready for the map or not. A lot of people do not approach maps this way (based on how my friends play) and get frustrated when they can't finish a map. Once I've broken down a map with them and asked them to work out their likely success on each difficult section, they quickly realise that they were being far too optimistic. It's a great way to manage expectation.

Share this post


Link to post
2 hours ago, rdwpa said:

 

I remember you reviewing this one. ITYTD and HNTR are the toughest two difficulties on that map. HMP and UV have more 'typical' monster compositions, and HNTR is the alternative 'lost soul and PE spam' mode. 

I need to revisit that map. Also I appreciate you chiming in on this map specifically. @Fonze will be happy about that too.

Share this post


Link to post
54 minutes ago, tmorrow said:

 

OMG! Mutiny, Stardate 20x7, Sunlust, Argh! A lot of vanishingly small probabilities to beat those maps. I think Wargames said it best: "the only winning move is not to play". Either that or IDDQD the entire way!

What you need are overpowered weapons:

 

Share this post


Link to post
12 minutes ago, Empyre said:

What you need are overpowered weapons:

 

What an ingenious solution to a very difficult problem!

Share this post


Link to post
26 minutes ago, Grazza said:

First exit. Total number of attempts? *shrug*

 

If I think something is possible, and I want to do it, I might set it aside for some time, but will never truly give up on it.

 

That's one great pacifist run Grazza! I'll have to add Henry to a long wad list of todo's. I had a quick pacifist run at map04 and got over the bridge but died to a revenant and pack guarding the 3rd room, that room is chock full of ugliness. I don't want to contemplate what the return journey is like after scoring the yellow key with the entire map riled up.

 

I've got a few maps that I haven't quite been able to do but that I think are still possible and will fire them up from time to time for another shot in the dark at them. On the other hand there are a lot of maps that are way beyond my skill level and I have little chance of beating even if I wanted to and it's good to be able to work that out as early as possible before investing a lot of time and effort.

Share this post


Link to post

How is this quantitative if N is rooted in a subjective guess about one's feelings?

Share this post


Link to post
6 minutes ago, Vorpal said:

How is this quantitative if N is rooted in a subjective guess about one's feelings

 

The quantitative part is in the probabilities assigned to subsections of the map. You are right that N is subjective, but after all it is the NOSE applied to your individual case. Also remember N is an up front guess on how many times you think you might be happy to play the map.  You can adjust/iterate on N it as you go so it becomes more tailored accurate to you and your feelings over more and more replays.

 

However, NOSE does not rely on N. If you really dislike N then leave it out. Work with the probabilities and compute your chances of finishing subsections of the map. Taking my map22 example from Super MAYhem 17, then based only on my probability assignments to the difficult sections gives P=1/48 of completing the map. Now use whatever better objective method you can come up with to decide whether it is worth going for or not.

Share this post


Link to post
11 minutes ago, rehelekretep said:

thats very clever - thanks for putting the effort in to that equation and post

 

Thank you @rehelekretep. I hope you can apply NOSE to your own play and increase your doom gaming happiness and satisfaction!

Share this post


Link to post
3 minutes ago, tmorrow said:

 

Thank you @rehelekretep. I hope you can apply NOSE to your own play and increase your doom gaming happiness and satisfaction!

Yes, it's an interesting algorithm, and anything that makes playing more enjoyable makes sense to me.

 

6 hours ago, tmorrow said:

 

Excellent algorithm. The only thing it doesn't cater for is an up front estimation of likelihood of success and possible disappointment/regret. You can add step 0 to address the disappointment though:

 

0. Go in with no expectation that you'll complete your goal. If you end up succeeding it's a bonus.

 

Heh - I forgot one BIG detail: I rarely save, or restart a level. Usually, if I can't beat the level outright, I'll switch to single-player coop mode, which lets me respawn at 100% health, without reloading. So, eventually, I *will* complete the map, and it's ok...even beneficial to get killed over and over. In this way, death is more of a powerup than anything else :)

Share this post


Link to post
6 minutes ago, kb1 said:

I'll switch to single-player coop mode, which lets me respawn at 100% health, without reloading.

 

Now that is a really cool way to play when you're not trying to go for a no death exit. You can even use that method to improve your overall  play by counting up your deaths during the run, then playing all over again and seeing if you do it with less deaths the next time. Thanks, I'm going to see how I can use this idea in my own play!

 

While there is no substitute for practice on particularly tough sections, your method allows you to derive the full pleasure of the map every time, a great way to play.

Share this post


Link to post
4 hours ago, tmorrow said:

 

Now that is a really cool way to play when you're not trying to go for a no death exit. You can even use that method to improve your overall  play by counting up your deaths during the run, then playing all over again and seeing if you do it with less deaths the next time. Thanks, I'm going to see how I can use this idea in my own play!

 

While there is no substitute for practice on particularly tough sections, your method allows you to derive the full pleasure of the map every time, a great way to play.

Yeah, it's pretty fun. Unfortunately, the port I use with that option is unreleased, but I'm pretty sure other ports, like PrBoom+ has a similar option - I think it's named "-nonet", or "-solonet" (someone please help with the exact spelling). Yeah, for actual competition, recording, streaming, etc., you've got to play proper, but for casual entertainment, it's hard to beat.

 

As soon as it stopped costing you 25 cents to play a video game, the concept of "game over" gets antiquated, in my opinion.

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×