My feelings on The Demon Rush
Game Page: Doesn't exist yet.
Verifier Responses
Decision: Reject
Reason: The runner optimized his run according to the in-game time, which doesn't count time spent in the menu. However, we can't use the in-game timer because it counts load times. Also, there were a couple of planning mistakes that cost a decent chunk of time.
BTW, I don't know why there were only 2 verifiers myself. I know there were a lot of people who watched the run when the runner put it up on youtube. If more of those people actually verified, maybe the run would have passed?
The run is available for download until June 1, just PM me for the link.
Verifier Responses
Quote:
Verifying run "Grand Theft Auto 4".
===============
Verdict: ACCEPT
===============
---
Cheating: no.
---
Timing: in-game-timer. [4:40:02]
---
Quality (capture):
i've used link (the verification link) also used ( for problem parts
"gta4-v-35.avi" - cannot play this part, that part was verified through youtube link ( "Grand Theft Auto 4 Speedrun (4:40:02) Part 15" starting 4:20 ending 9:00)
"gta4-v-34.avi" - audio level here is noticably lower comparing to other parts (also noticed within corresponding youtube video) ( 2:30)
"gta4-v-54.avi - at 2:56 there are several seconds of encoding bugs (health-ammo gauge) // not noticed in corresponding youtube ( "Grand Theft Auto 4 Speedrun (4:40:02) Part 27" 2:51)
hope there will be HQ-version.
---
Quality (gameplay):
1. Planning: Whole run is very well planned.
1.1 Taxis are used to save time needed for driving.
1.2 Short-cuts are used heavily while driving (when taxis cannot be used).
1.3 Short-cuts/time-savers for mission-start or mission-finish are used.
1.4 Cut-scenes are skipped (not all, some scenes cannot be skipped)
1.5 Phone conversations are skipped.
1.6 "Vehicle Density" is manipulated (decreased before driving parts, increased before looking-for-taxi parts)
1.7 Missions not-needed for finishing the game are not played through.
1.8 During "Loose your wanted level" parts vehicle is moving towards final destination (if taxi cannot be used afterwards)
1.9 "Revenge" branch is chosen
2. Execution/playing: Excellent
2.1 Luck manipulation is used while looking-for-taxi
2.2 Driving/riding/flying is fast. Nice turns, quick landings, max speed.
2.3 Running and sprinting.
2.4 A lot of head-shots while shooting.
2.5 Shooting while moving or while driving.
Of course there are some imperfections during the run, which cause seconds or sometimes fractions of a second. These are totally acceptable taking into account whole run's time. (No significant time losses).
---
Consistency:
Since the run is multi-segment one there were things-to-check between saves.
1. Save slot. After-save's load should go from same slot. Load which takes place after auto-save should go from auto-save slot.
2. Load after auto-save. Nico should apper in save-house which is closest to auto-save place.
3. Nico should wear the same clothes.
4. Money amount.
5. Health/Armor. Gauges with no numerical representation are used in game.
6. Weapons/Ammo. Weapons are checked, ammo is checked where possible.
---
Questions:
1. Game patch/version ? (1.0.4.0 ?)
2. part06: "CRIME AND PUNISHMENT": Is first van always wrong one?
3. Is segment amount correct? ( part22 = auto-save takes place, but no load, part25 = 3 saves in-a-row )
---
Biggest time loss: part37: "THREE LEAF CLOVER": looking for taxi after loosing wanted level.
---
Surely it would be nice to hear comments from runner, if he decides to do audio commentary.
===============
Verdict: ACCEPT
===============
---
Cheating: no.
---
Timing: in-game-timer. [4:40:02]
---
Quality (capture):
i've used link (the verification link) also used ( for problem parts
"gta4-v-35.avi" - cannot play this part, that part was verified through youtube link ( "Grand Theft Auto 4 Speedrun (4:40:02) Part 15" starting 4:20 ending 9:00)
"gta4-v-34.avi" - audio level here is noticably lower comparing to other parts (also noticed within corresponding youtube video) ( 2:30)
"gta4-v-54.avi - at 2:56 there are several seconds of encoding bugs (health-ammo gauge) // not noticed in corresponding youtube ( "Grand Theft Auto 4 Speedrun (4:40:02) Part 27" 2:51)
hope there will be HQ-version.
---
Quality (gameplay):
1. Planning: Whole run is very well planned.
1.1 Taxis are used to save time needed for driving.
1.2 Short-cuts are used heavily while driving (when taxis cannot be used).
1.3 Short-cuts/time-savers for mission-start or mission-finish are used.
1.4 Cut-scenes are skipped (not all, some scenes cannot be skipped)
1.5 Phone conversations are skipped.
1.6 "Vehicle Density" is manipulated (decreased before driving parts, increased before looking-for-taxi parts)
1.7 Missions not-needed for finishing the game are not played through.
1.8 During "Loose your wanted level" parts vehicle is moving towards final destination (if taxi cannot be used afterwards)
1.9 "Revenge" branch is chosen
2. Execution/playing: Excellent
2.1 Luck manipulation is used while looking-for-taxi
2.2 Driving/riding/flying is fast. Nice turns, quick landings, max speed.
2.3 Running and sprinting.
2.4 A lot of head-shots while shooting.
2.5 Shooting while moving or while driving.
Of course there are some imperfections during the run, which cause seconds or sometimes fractions of a second. These are totally acceptable taking into account whole run's time. (No significant time losses).
---
Consistency:
Since the run is multi-segment one there were things-to-check between saves.
1. Save slot. After-save's load should go from same slot. Load which takes place after auto-save should go from auto-save slot.
2. Load after auto-save. Nico should apper in save-house which is closest to auto-save place.
3. Nico should wear the same clothes.
4. Money amount.
5. Health/Armor. Gauges with no numerical representation are used in game.
6. Weapons/Ammo. Weapons are checked, ammo is checked where possible.
---
Questions:
1. Game patch/version ? (1.0.4.0 ?)
2. part06: "CRIME AND PUNISHMENT": Is first van always wrong one?
3. Is segment amount correct? ( part22 = auto-save takes place, but no load, part25 = 3 saves in-a-row )
---
Biggest time loss: part37: "THREE LEAF CLOVER": looking for taxi after loosing wanted level.
---
Surely it would be nice to hear comments from runner, if he decides to do audio commentary.
Quote:
GTA 4 Verification
There is a lot to be said so I'll actually structure my comments with subtitles and try to be succinct rather than rambling like I normally do.
Timer Issues
There is a major problem that I think will make it impossible for SDA to accept this run, so I'll bring it up first and in detail. The runner clearly intended the in-game timer to be used. However, the in-game timer partly times loading screens.
I did some tests, consisting of checking the in-game timer just before and after a taxi ride skip and comparing with timing stuff with a stopwatch. In a typical taxi ride on my system, there's some 10-20 seconds (real time) of loading screens, of which some 3-10 seconds is included on the in-game timer, with more time passing on the in-game timer the longer the loading screen is. In other words, while the timer does not time entire loading screens, it times a significant amount of them and that amount depends on loading screen length. Presumably it will be less on faster computers, too.
Also, this run is optimised for the in-game timer, not for real time, so timing this run manually would mean it would be horribly suboptimal due to all the unnecessary time spent in menus. As such, I think our only option is to reject. All the same, I will give a full review of the gameplay and possible improvements in the hope that the runner will do another run.
Summary of actual run quality
Firstly, the fact that GTA4 has autosaves and taxi warps makes it a very different game for speedrunning to its predecessors. Before, segments were long and route planning revolved around finishing missions near the start of other missions. Now, the optimal strategy once taxis are available is to have only one or two missions per segment to allow optimal taxi manipulation. The runner does this.
Short segment length notwithstanding, execution in this run is still damn impressive. Even at its worst it is decent; at its best, in some of the shooting missions, it's insane. But as happens frutratingly often, the run is let down by a significant handful of major planning errors, some of them really basic ones that could've been avoided just by doing basic testing or flicking through the walkthrough at www.gta4.net, and almost all of them things that wouldn't have happened if the runner had simply read the threads at SDA before doing the run. In all fairness there's a few clever tricks the runner came up with himself, though.
I'll give details of all the improvements I've found below for the runner (or anyone else) to use if he does a rerun. The only major issues I haven't really tested are which missions are necessary (a lot of work to test and I think I'll just trust the runner hasn't screwed up here) and how to minimise the time wasted waiting for plot-advancing phone calls, which eats huge amounts of time in the final segments. I suspect, though I have not tested, that rather than devoting multiple segments to phone calls at the end of the run, it would have saved a minute or so if the runner had spread them out earlier in the run at the beginning of segments before taking missions. Then some of the time spent getting into taxis and travelling to mission markers would overlap with time spent waiting for phone calls, saving time over waiting for them with nothing else to do.
Mistakes
Really big basic mistakes that together I think warrant rejection are underlined.
1) I found a neat trick right after the first mission (the only trick I found that hadn't
already been pointed out months ago on the forum, I later discovered). If you jack a car before finishing The Cousins Bellick and park it outside the hideout, then later leave the hideout by entering the car instead of simply running out of the door, you'll skip the 15-second tutorial about hot dogs entirely for a net saving of a few seconds.
2) The runner doesn't use the trick, posted on the SDA forums, that lets you do First Date before Bleed Out (without having to pick Roman up from the hospital). He also needlessly triggers the cutscene with Darden's thugs beating up Roman instead of just shooting them from the car.
3) The runner should've just done Easy Fare in segment 2, then had a resaving segment to advance time to get the Vlad mission that unlocks taxis so he could use them for Jamaican Heat and Concrete Jungle, which would save several tens of seconds.
4) On Hung Out To Dry, the van can be stopped earlier by shooting out one of its front tyres as it goes into the first turn to spin it.
5) On clean getaway, the runner should've cleaned the car at the pay 'n' spray next to the garage he has to deliver it to, instead of at the car wash. This was a pointless 30 second detour.
6) In Uncle Vlad, the runner chases Vlad out of the back door and runs around the building to get back to his car instead of simply turning around and going out the front door instead, which is silly but only costs 5 seconds or so.
7) Do You Have Protection is absolutely riddled with mistakes. Firstly, the runner should've manually called Roman to unlock the mission, instead of waiting for his call. Secondly, the runner drives huge distances instead of using a taxi; my only explanation for this is that he hadn't realised yet that you can use taxis when someone is following you, which is a pretty huge and basic mistake. Thirdly, the runner should've run behind the guys you have to shoot in the leg and shot the guy who has to give money to Dimitri towards Dimitri instead of away from him to save a second or two. Fourthly, as soon as the money is handed over, the runner should've skipped the rest of the lengthy dialogue by killing the remaining porn guys and leaving, which prompts Dimitri to say 'This is a mess. Let's get out of here.' and run out of the building after you, instead of walking out slowly. Together these mistakes add up to almost 2 minutes wasted on a single mission.
8) On Logging On, the runner could've saved a second by shooting to scare the receptionist BEFORE talking to her to avoid a second of dialogue.
9) The runner squanders some 15-20 seconds or so on Escuela Of The Streets by, again, driving instead of using a taxi when he has a companion. (After this mission, the runner always correctly uses taxis whenever possible.)
10) The autosave at the end of Luck Of The Irish is weird in that, unlike every other mission, it happens after the post-mission phone call instead of when the mission actually ends. However, if you jump over the edge of the balcony so you're falling when the mission ends and hence can't immediately make the call, the call is postponed and the save happens immediately, saving 4 seconds.
11) The runner fails to skip several cutscenes on Easy As Can Be and Call And Collect, probably because he didn't realise they were skippable (they need to be skipped with enter instead of spacebar). This costs a minute. (Cutscenes that need to be skipped with enter later in the run are.)
12) The runner shoots Goldberg and leaves through the front entrance on Final Interview, meaning he has to deal with a shootout and escape from police. It is about a minute faster to silently knife Goldberg and leave through the window. This solution to the mission was posted about on the forum and can be found in any walkthrough.
13) On Late Checkout, it would've been a good 10 seconds quicker to leave the roof via the outside elevator instead of the inside one. Even more time could almost certainly have been saved by using one of the multiple 'damageless falling' tricks posted on YouTube or on the forum to jump off the roof without dying.
14) Killing Derrick in Blood Brothers is a mistake. As any walkthrough or basic testing could tell you, it means you have to drive to a much more distant cemetery on Undertaker, and costs 40 seconds.
15) There's 5-10 seconds to be saved on Dining Out by using a rocket to kill the manager instead of running into the office and back again.
16) Buoys Ahoy can be completed over 45 seconds earlier without adding any difficulty by simply shooting everyone on the boat instead of shooting the boat until it eventually explodes. I imagine a full 60 seconds would be savable by blowing up the boat with a rocket at the very start of the level, but rocket trajectories are so random that would take some luck and I wasn't able to pull it off in testing.
After all that criticism, I feel the need to point out a couple of moments that especially impressed me (not the only ones, just a couple that spring to mind).
1) Luck-manipulating a high-suspension vehicle on Waste Not Wants Knots in order to take an otherwise untravesible route was a great idea that could easily have been missed.
2) The shooting on Entourage was some of the fastest aiming at heads I've ever seen.
Conclusion
I'm going to have to reject this. Firstly, the in-game timer can't be used because it times a significant portion of the loading screens, which means the run would have to be timed manually, which means that all the time spent on the map screen (much of it unnecessary, or spent having a quick breather after placing a waypoint before going back to the game) would end up costing several minutes.
Even if that issue didn't exist, I would reject for run quality. While the execution is impressive throughout and there are a handful of nice tricks in there, there are just too many basic mistakes, together costing well over 5 minutes, resulting from a lack of simple knowledge that could've been easily picked up on the forum, through walkthroughs, or through the most simple testing. On top of that, there are other, less obvious or less serious mistakes costing even more time – and there's bound to be others I've missed not included in my list above.
All that said, I'll emphasise once again that the execution is tight and the runner is clearly very skilled, it's just that this run was a bit premature and he dove into it without doing anywhere near enough testing and without reading the stuff that had already been posted on SDA and elsewhere that would've helped him shave many minutes off his time. I really hope he isn't discouraged and that he'll do another run improving all the mistakes we're able to find and submit it.
There is a lot to be said so I'll actually structure my comments with subtitles and try to be succinct rather than rambling like I normally do.
Timer Issues
There is a major problem that I think will make it impossible for SDA to accept this run, so I'll bring it up first and in detail. The runner clearly intended the in-game timer to be used. However, the in-game timer partly times loading screens.
I did some tests, consisting of checking the in-game timer just before and after a taxi ride skip and comparing with timing stuff with a stopwatch. In a typical taxi ride on my system, there's some 10-20 seconds (real time) of loading screens, of which some 3-10 seconds is included on the in-game timer, with more time passing on the in-game timer the longer the loading screen is. In other words, while the timer does not time entire loading screens, it times a significant amount of them and that amount depends on loading screen length. Presumably it will be less on faster computers, too.
Also, this run is optimised for the in-game timer, not for real time, so timing this run manually would mean it would be horribly suboptimal due to all the unnecessary time spent in menus. As such, I think our only option is to reject. All the same, I will give a full review of the gameplay and possible improvements in the hope that the runner will do another run.
Summary of actual run quality
Firstly, the fact that GTA4 has autosaves and taxi warps makes it a very different game for speedrunning to its predecessors. Before, segments were long and route planning revolved around finishing missions near the start of other missions. Now, the optimal strategy once taxis are available is to have only one or two missions per segment to allow optimal taxi manipulation. The runner does this.
Short segment length notwithstanding, execution in this run is still damn impressive. Even at its worst it is decent; at its best, in some of the shooting missions, it's insane. But as happens frutratingly often, the run is let down by a significant handful of major planning errors, some of them really basic ones that could've been avoided just by doing basic testing or flicking through the walkthrough at www.gta4.net, and almost all of them things that wouldn't have happened if the runner had simply read the threads at SDA before doing the run. In all fairness there's a few clever tricks the runner came up with himself, though.
I'll give details of all the improvements I've found below for the runner (or anyone else) to use if he does a rerun. The only major issues I haven't really tested are which missions are necessary (a lot of work to test and I think I'll just trust the runner hasn't screwed up here) and how to minimise the time wasted waiting for plot-advancing phone calls, which eats huge amounts of time in the final segments. I suspect, though I have not tested, that rather than devoting multiple segments to phone calls at the end of the run, it would have saved a minute or so if the runner had spread them out earlier in the run at the beginning of segments before taking missions. Then some of the time spent getting into taxis and travelling to mission markers would overlap with time spent waiting for phone calls, saving time over waiting for them with nothing else to do.
Mistakes
Really big basic mistakes that together I think warrant rejection are underlined.
1) I found a neat trick right after the first mission (the only trick I found that hadn't
already been pointed out months ago on the forum, I later discovered). If you jack a car before finishing The Cousins Bellick and park it outside the hideout, then later leave the hideout by entering the car instead of simply running out of the door, you'll skip the 15-second tutorial about hot dogs entirely for a net saving of a few seconds.
2) The runner doesn't use the trick, posted on the SDA forums, that lets you do First Date before Bleed Out (without having to pick Roman up from the hospital). He also needlessly triggers the cutscene with Darden's thugs beating up Roman instead of just shooting them from the car.
3) The runner should've just done Easy Fare in segment 2, then had a resaving segment to advance time to get the Vlad mission that unlocks taxis so he could use them for Jamaican Heat and Concrete Jungle, which would save several tens of seconds.
4) On Hung Out To Dry, the van can be stopped earlier by shooting out one of its front tyres as it goes into the first turn to spin it.
5) On clean getaway, the runner should've cleaned the car at the pay 'n' spray next to the garage he has to deliver it to, instead of at the car wash. This was a pointless 30 second detour.
6) In Uncle Vlad, the runner chases Vlad out of the back door and runs around the building to get back to his car instead of simply turning around and going out the front door instead, which is silly but only costs 5 seconds or so.
7) Do You Have Protection is absolutely riddled with mistakes. Firstly, the runner should've manually called Roman to unlock the mission, instead of waiting for his call. Secondly, the runner drives huge distances instead of using a taxi; my only explanation for this is that he hadn't realised yet that you can use taxis when someone is following you, which is a pretty huge and basic mistake. Thirdly, the runner should've run behind the guys you have to shoot in the leg and shot the guy who has to give money to Dimitri towards Dimitri instead of away from him to save a second or two. Fourthly, as soon as the money is handed over, the runner should've skipped the rest of the lengthy dialogue by killing the remaining porn guys and leaving, which prompts Dimitri to say 'This is a mess. Let's get out of here.' and run out of the building after you, instead of walking out slowly. Together these mistakes add up to almost 2 minutes wasted on a single mission.
8) On Logging On, the runner could've saved a second by shooting to scare the receptionist BEFORE talking to her to avoid a second of dialogue.
9) The runner squanders some 15-20 seconds or so on Escuela Of The Streets by, again, driving instead of using a taxi when he has a companion. (After this mission, the runner always correctly uses taxis whenever possible.)
10) The autosave at the end of Luck Of The Irish is weird in that, unlike every other mission, it happens after the post-mission phone call instead of when the mission actually ends. However, if you jump over the edge of the balcony so you're falling when the mission ends and hence can't immediately make the call, the call is postponed and the save happens immediately, saving 4 seconds.
11) The runner fails to skip several cutscenes on Easy As Can Be and Call And Collect, probably because he didn't realise they were skippable (they need to be skipped with enter instead of spacebar). This costs a minute. (Cutscenes that need to be skipped with enter later in the run are.)
12) The runner shoots Goldberg and leaves through the front entrance on Final Interview, meaning he has to deal with a shootout and escape from police. It is about a minute faster to silently knife Goldberg and leave through the window. This solution to the mission was posted about on the forum and can be found in any walkthrough.
13) On Late Checkout, it would've been a good 10 seconds quicker to leave the roof via the outside elevator instead of the inside one. Even more time could almost certainly have been saved by using one of the multiple 'damageless falling' tricks posted on YouTube or on the forum to jump off the roof without dying.
14) Killing Derrick in Blood Brothers is a mistake. As any walkthrough or basic testing could tell you, it means you have to drive to a much more distant cemetery on Undertaker, and costs 40 seconds.
15) There's 5-10 seconds to be saved on Dining Out by using a rocket to kill the manager instead of running into the office and back again.
16) Buoys Ahoy can be completed over 45 seconds earlier without adding any difficulty by simply shooting everyone on the boat instead of shooting the boat until it eventually explodes. I imagine a full 60 seconds would be savable by blowing up the boat with a rocket at the very start of the level, but rocket trajectories are so random that would take some luck and I wasn't able to pull it off in testing.
After all that criticism, I feel the need to point out a couple of moments that especially impressed me (not the only ones, just a couple that spring to mind).
1) Luck-manipulating a high-suspension vehicle on Waste Not Wants Knots in order to take an otherwise untravesible route was a great idea that could easily have been missed.
2) The shooting on Entourage was some of the fastest aiming at heads I've ever seen.
Conclusion
I'm going to have to reject this. Firstly, the in-game timer can't be used because it times a significant portion of the loading screens, which means the run would have to be timed manually, which means that all the time spent on the map screen (much of it unnecessary, or spent having a quick breather after placing a waypoint before going back to the game) would end up costing several minutes.
Even if that issue didn't exist, I would reject for run quality. While the execution is impressive throughout and there are a handful of nice tricks in there, there are just too many basic mistakes, together costing well over 5 minutes, resulting from a lack of simple knowledge that could've been easily picked up on the forum, through walkthroughs, or through the most simple testing. On top of that, there are other, less obvious or less serious mistakes costing even more time – and there's bound to be others I've missed not included in my list above.
All that said, I'll emphasise once again that the execution is tight and the runner is clearly very skilled, it's just that this run was a bit premature and he dove into it without doing anywhere near enough testing and without reading the stuff that had already been posted on SDA and elsewhere that would've helped him shave many minutes off his time. I really hope he isn't discouraged and that he'll do another run improving all the mistakes we're able to find and submit it.
Decision: Reject
Reason: The runner optimized his run according to the in-game time, which doesn't count time spent in the menu. However, we can't use the in-game timer because it counts load times. Also, there were a couple of planning mistakes that cost a decent chunk of time.
BTW, I don't know why there were only 2 verifiers myself. I know there were a lot of people who watched the run when the runner put it up on youtube. If more of those people actually verified, maybe the run would have passed?
The run is available for download until June 1, just PM me for the link.
Thread title: