HotS replays, and reconnect system, work by scripting out all the actions players make in the duration of the game. The replay file just runs the scripts to have all those actions happen again, so watching old replays requires a version match to the file so the values in the game remain the same.
Since the file re-recreates the same actions, it’d be neat if a program/viewer allowed some edits to values used by the script. The correct/original information would still be needed, but a drop-down menu could compare the original stats compared to the edited ones. (for showing more or less damage/healing over the same actions)
So picture if we could run a replay as "what if [Hero] picked ‘y’ talent instead of ‘x’ talent, and then the end screen shows instances of invalid differences compared to the original (ie, where an armor buff prevented otherwise lethal damage, extra healing numbers in that instance ‘saved’ an ally, extra damage boost secures a kill earlier, etc)
So the program would essentially duplicate a replay file, has specific variables that can be changed (such as talent tracking) and then replay the whole thing twice, and compare the end results; only noting that some instances have ‘invalid’ actions (such a hero dying sooner due to changes) rather than crash or glitch out the reply. I guess It’d be like Ghosting in Mario Cart, where the game saves your best (or previous) run and has the Ghost do the same actions, so you can compare one performance against another.
How should it solve kill vs no kill situations?
original - player 1 gets killed by player 2 in lane and is 30 seconds in respawn queue
alternative - player 1 survives.
what should he be doing during his 30 seconds of non-respawn queue/travel time? what should player 2 do if player 1 taps and returns to lane 4 seconds later instead of having uncontested lane?
original - player 1 lives and does a 5 man mosh followed by a teamwipe
alternative - player 1 gets killed 10 seconds earlier and simply is not there.
solution?
just ignoring a hero being dead due to changed circumstances and still generating meaningless numbers seems rather pointless
not even going into movespeed/displacement changes that would generate bodyblock vs free path situations
yes, all of that becomes a huge issue if it were a straight replay. you cant just expect a replay to do that kind of thing.
i suspect what you would really need is some sort of integration between an AI game and replay system. which could give you the ability to do a live replay using AI vs watching a replay only. in the case of AI when choosing a different talent if you selected a specific point on the time line, say level 10 when you get ults, and then start playing from there to see what will happen choosing the other ult. maybe you want to see how the AI would have handled Stagedive vs Mosh in the situation you were in. A live AI game could handle the changes to talents like this.
i dont expect this is something they are going to implement any time soon, but it would be interesting.
It just makes a timestamp note of where something would be different (ie hero lives or hero dies and thus can’t do the same actions it would be expected to do at that point) It’d need a similar exception in the event an ability cooldown wouldn’t be up due to talent changes (second charge or cdr) and it simply doesn’t do the action.
Replay ghosting would need the actions to remain the same, only the variables on the numbers for those actions to shift; that’s why the adjustments would have to be specific and not just everything. Taking a different heroic would require different actions and be significantly more complicated.
The idea is to compare simpler adjustments to simulate if there was more value with how a player played compared to what they thought they’d get out of it.
for objective numbers breakdown you can use the target dummy in try mode, for talents it just doesn’t make too much sense as circumstances simply aren’t comparable when a talent doesn’t do just x amount numbers change.
hanzo W quest adds projectiles that have to be evaluated
picking or not picking calamity greatly impacts the use of teleport
talents that add/enhance slows or enhance movespeed mess up the whole scripted positioning
garrosh lv1/tychus lv7 grenade changes range of throw and AoE/knockback
… and couple hundred more cases that are not covered by just “04:56 - no kill here”
when was the last time you bothered to watch a replay?
when was the last time someone else mentioned they were going to watch a replay?
When replay topics come up, the usual response is that the system is clunky, they can’t rewind and so on and so forth that of the few that have tried it, they convince themselves to not bother.
Part of a long standing issue with HotS is that players lack a sense of self-evaluation, a sense of experimentation, and a sense for alternative perspectives.
The basic premise is to try to take the unusual aspect of how HotS does replays and had a sense of control to something people ‘know’ and present them with something they might not; if a player could adjust a few base values to hp/aa and run the replay again, they can look at the direct impact of one patch change to another rather than assuming nothing happened.
Similarly, a number of talents don’t see as much variety in their choice so if people can plug in the switch and find out that how they played may have favored a different pick, then they may be more open to situations they didn’t consider in favor of just prattling on about ‘knowing’ the consequence of the choice at hand, or just following the ‘meta’ because its the meta.
Yes, context generally matters more than just “oh you dealt a little extra here, or didn’t secure the kill here” but a hefty chunk of players reject the aspect of context that isn’t the tl;dr.
If the game isn’t going to get a map making editor or workshop, then editable replay value scrips may be the closest thing possible with the way the game works now.
If people can’t “learn from ai” then they certainly only get miniscule data points from a test dummy because the dummy reacts even less than the ai would. If the numbers the dummy provides are useful, then magical guess what, the numbers provided by a replay ghost would also be useful.
A simpler tool that could provide critical data for the player might be something to the effect of tracking talent specific “bonus” values.
More data presented to the player is always good, and there’s little reason to remove things like with “damage soaked.” A comprehensive summary after the game could allow for meaningful reflection on talent choices. I would love to see how much CC time, uptime of abilities, breakdowns of damage and healing mitigated, all specifically connected to their respective talents and abilities.
Obviously what you’re suggesting could provide even more insight. However, the implementation seems very challenging.
I like that idea. Infact it’d be neat if a replay would have some sort of “what if” pilot scenario where you can toggle and control the Heroes on there for the brief moment, but upon doing so, you have to exit the replay and open up again for it to revert the replay to the way it originally was. That way instead of players saying “x should’ve done y”, you’d get to do it inside the replay and see if “y” would’ve worked. After you log out of that specific replay, that replay will always revert to the original match of what it was. I can imagine this being cool on WC 3 Reforged. I hear a lot of commentators or players back-seat driving and saying “he should’ve sent this to flank it” and if they tried to do it themselves on the replay, it would be impossible too. This would also open up to fixing mistakes and learning, and in other instances, practicing vsing top tier players depending on the game, but the AI in the replay will just do exactly what the opponents did in the match
Edit: Actually, it’d probably be weird for the AI in the replay lol. Imagine going here or there, but the opponent AI just keeps casting spells or AAing and going to that location where the hero was originally meant to be.
yea, that ties a bit into something similar mogorath was suggestion with AI additions.
If that were the case, I think the ‘edit’ would need to only work with a few moments of a replay as a snapshot of time to establish a reference, and then the ai parameters kick in from there.
Instead of modifying values used in an identical script, that would be modifying some of the actions taken by the script.
February, when I got nubkeks and Kendric humiliated in brawl
Why the heck would someone announce their replay watching like who cares?
Enhanced sandbox mode or custom games with preset AI builds would be better here, as mentioned earlier picking a talent might realistically prevent situations from the base replay starting with lv1.
“look this talent got me a kill in a situation that is literally impossible in any live game.” is no useful information to obtain if you ask me.
I somewhat see your point, but if few people watch replays, how many would juggle +/- x% changes on patch day just to not play 100 games on the hero to noticeably benefit from +2% overall winrate?
As mentioned above, custom games would be a better tool here. Giving a teleport ming replay all orbs talents does provide zero useful information. Better to run a whole new match with the same setup, same builds of team and enemy if appropriate and see how it turns out.
It’s obviously not an rts we’re playing where certain situations are probable or to be expected repeatedly over multiple matches.
I doubt those exact players would sit and run the same replay 3 times to get some numbers that may or may not be of any value
Sure, but magical guess what, the target dummy is already sitting and waiting.
The first question isn’t leading enough, the second ones the tell, the following paragraph has the answer for the rhetorical.
Generally if people only see lines and respond to the lines as they see them, they don’t consider the progression or cycular value of context as a whole.
People that think themselves above hypothetical scenarios don’t bother to think through the points of thought experiments because they already have the answer they think they want.
The whole point of a suggestion like this would be to demonstrate that there are things people don’t consider. Even if the idea of “oh well they won’t use it” were a complete conclusion, the typical instances of why people ‘think’ something is better than it ‘could’ be is because the idea was reinforced by something as is, such as hotslogs or heroesprofile where the successful/popular picks are taken because “they’re successful”
Replay ghosting would be an alternative to post comparable figures faster than it would take to play a whole game through; that’s part of the whole point of a replay. Similarly, the conditions from one game to another will be different whether it be compositions, maps, or a myriad of other stuff (things you’d consider to be stuff why they wouldn’t want to do the test, rather than why they could)
Phrased different, it’s a way of having a controlled experiment, and having a fast forward button.
The basic gist, as noted by the answering a rhetorical question, is that you’re too busy being content with what you already have to consider the value of what you don’t consider.
And that’s part of why players don’t bother with replays; they don’t have the observational awareness to care about much else then thinking themselves a super leader carrying the way to victory by being a lemming.
