• Content count

  • Joined

  • Last visited

About tomshackell

  • Rank
    Pop up Target

Contact Methods

  • Website URL

Profile Information

  • Location
    York, UK

Profile Fields

  • Platform
    PC gamer
  1. AI makes or breaks a game

    Totally agree on enemy AI being more important than friendly AI. Personally I will probably only play GB coop, so friendly AI is totally a non-issue. I also agree with Jsonedecker's earlier post where he said that the difficult thing is making the AI understand their environment in a believable way. I was trying to build a mission in Arma 2 the other day, but in the end I got disappointed and gave up. I was trying to make a simple night ops mission; rescue some captured POWs and then exit. However the enemy AI were just not playing ball. Apparently they were totally blind at night, and would not see you even at point blank range. And would never fire back, even at the (very obvious) muzzle flashes. Needless to say this did not make for a convincing mission. Another classic example is OFP 2. They made a big song and dance about how the AI had "Play books" and how it would flank you and all sorts. That might have been true but they didn't get the basics right. They didn't get the right vision model for them. They had 360 degree vision. They could spot you from half a mile away, when you've crawled the whole mission, and haven't yet fired a shot. They also had X-ray vision that could see through the grass and bushes and all sorts. For me AI really does make or break a game. If I could have just one feature that was perfect in GB it would be to have an awesome enemy AI. If you've got that everything else falls into place for me.
  2. Fire selector

    I voted for option 2. My main objection to option 1 is that I never use burst mode, so I wouldn't want 2 out of four cycles to take me to a mode I never use
  3. Randomised patrols: Part 2

    Hello, A while back I posted giving some experience with randomising in operation flashpoint: Well recently I modded RavenShield to do a similar thing, to make the enemies move in random patrols. I thought I'd post my experiences with it, in case anyone was interested. And to once again blow the "Pro Randomisation" trumpet Our group has been playing RvS cooperatively for years ... and that was the problem. RvS spawns enemies somewhat randomly, but we'd learnt where all the spawn points were. This made the game far too predictable for us. However recently I discovered an unofficial SDK for RavenShield and thought it was time for some random patrols. What I did was pretty simple. The map is filled with 'path nodes' for controlling the AI movement, so I had the tangos do the following: Wait a random period of time, between 5 and 90 seconds. Pick 4 path nodes at random from all the nodes on the map, and then take the closest node of the 4. Walk towards that chosen path node. When they reach the target path node, go back to step 1. At first I had them just pick a node at random from the whole map, but that meant that they tended to criss-cross the map too much. Taking the closest of 4 tends to keep them in the same area, but occasionally they will wander further. There were a few other tweaks needed to prevent clustering, and to ensure hostages were still being guarded - nothing complicated. The effect was enormous, suddenly rear guard became a real position ... something that mattered ... previously it was something that someone got assigned ... but they'd largely ignore the duty and just face forwards with everyone else. We quickly found that if we did that with random patrols it wasn't long before we had all been shot in the back. It also meant we never took a room in quite the same way twice. Different numbers of enemies, different positions. We could play the same room 20 times and get 20 different scenarios. Much much more interesting. Once again such a simple mod has totally changed the way we play. We're playing more as a team, with better coordination and keeping 360' cover at all times. We're also not ploughing through levels like we were. There is no "magic path" through the level any more, and every play through presents a unique challenge. Once again, randomisation has brought our gaming to a new level.
  4. News or Updates?

    Not much to post about, but I'm still sitting and waiting to see what's happening. Ground branch hasn't been forgotten, after all Ground Branch is our "Last, best hope for ... something other than Call of Duty IX"
  5. Wot would you like to see randomized?

    Yes, I did find that randomised objectives is more tricky. The reason being that the player is extremely good at exploiting the AI's weaknesses, but the AI is not good at exploiting a player's weaknesses. For example in OFP the player generally has a huge advantage at long range. As a result I've found you have to structure the missions quite carefully so that the player can't simply snipe off all the enemy. Random objectives makes that much harder to achieve.
  6. Wot would you like to see randomized?

    Yes, I found that random spawn positions by themselves isn't enough. Raven Shield had random spawn positions, however, you very quickly learnt where they all were. As a result, you couldn't say for sure whether an enemy was in a particular place, but you knew exactly where to look. However, random patrols I've found are very effective, even if the path itself isn't random. The reason is that you can have lots of patrols all with criss crossing paths. If you randomly pick which patrols will appear then you don't know which path a particular patrol is on. Combine that will placing the patrol at a random start point along that patrol route, and throw in random delays along the patrol and you end up with a situation that is totally random to all intents and purposes. Truly random paths would be very cool, but also harder to implement
  7. Enemies tend to look exactly the same

    Personally I would say having very varied enemies is a "nice to have", but there are lots of other features I'd like to see first
  8. Wot would you like to see randomized?

    Just as an update to this thread, my friends and I have been playing the original Operation Flashpoint (OFP) recently. OFP doesn't come with a huge number of missions suitable for small team coop so I've been making some. One of the really nice things about OFP is that it has a pretty powerful scripting language, the language is archaic and quirky but if you can put up with that then you can do a lot of stuff in it. So I created some missions that have random patrols in. The idea is pretty simple: I place 8-12 enemy groups going round the mission area in circular patrols. When the mission runs, the script first picks 4 or 5 enemy groups to play with, and deletes all the others. For each of the chosen groups it chooses a random waypoint along their circular patrol route, and places them there. Simple concept but we've found the effect is incredible. You don't know which groups will be present and which won't, and you have no idea at what point they will be along their (large) circular patrol route. This means all areas are potential threat areas, and you never know where the enemy will appear from. This had a massive impact on our gameplay. Instantly we found we had to observe our surroundings very carefully. Running around haphazardly through the mission area got us all killed very quickly. We also found that teamwork became essential. You need to play as a coordinated unit, covering each others movements and maintaining 360 awareness. Knowing when to open fire and when to maneuver to a better position became far more critical. As did choosing how and when to engagement, for example "we shouldn't attack from here, we don't have a good view of our flank". We also found we had to change our plans a lot, something that seemed like a good idea at first could quickly seem like a bad idea when an enemy patrol turned up in exactly the wrong place. Mission tactics became far more dynamic and interesting. Lastly we found that random patrols makes the game much *much* harder. A 400x400 meter mission area, with about 10-15 enemies on it suddenly became a real challenge. This was true even though the AI in OFP is actually quite weak in many ways. You couldn't just learn the enemy positions, and there was no "magic path" through the mission. Fortunately the random patrols also meant tweaking the difficulty was really easy. Simply changing the number of groups that the script picked had a big effect on difficulty and allowed us to come up with something that really suited our group size well. All in all I'd say that personally speaking random patrols have had a bigger effect on our gameplay than any other feature. It's really improved our gaming experience, immensely.
  9. Acceleration-based aiming system

    Happy to do you one in Java if you'd like ... well for the reading mouse coordinates and displaying stuff at least
  10. Hostages!

    Actually I disagree, I think good random preprogrammed behaviour is better than mission scripting. The reason being that a script is unpredictable exactly once. The second time you play the mission you get someone to stand in the door because you know the hostage is going to try and leg it .. or you shoot a hostage on sight because you know he was actually a terrorist in disguise ... or whatever. You can do some amazingly complicated things with mission scripting but personally I prefer missions that are very random with very little pre-scripted behaviour. That way you never know what to expect. As far as the hostages AI goes, personally I'd say it's not a priority. Especially as GB/SG will not be focussed on hostage rescue I think the time would be spent elsewhere. Obviously if you have hostages they need to do the obvious things. They need to follow the player when asked and maybe sometimes getting scared and running off, and perhaps utter the odd "are you here to rescue us?", "My god I was so scared" comments for a bit of immersion.
  11. The most important things

    Indeed, I think everyone here would agree on that
  12. Aim convergence

    My biggest complaint with ArmA with regards to weapon modelling was that the movement always felt really weird. The walking felt weird, the way the weapons handled felt weird. If the sights are always aligned then I guess, from the point of my original question, ArmA has neither linear nor non-linear convergence. Since it actually has no convergence at all. Having said that I can't believe the sniper rifles would have behaved that way in ArmA (otherwise you'd have "insta-aim") ...
  13. The most important things

    There have been lots of suggestions for features in SG/GB. However, something I'd be really interested to know is what things people thought were the most important? What are people's "must have" features that they'd be really excited about seeing in game? I guess it'd be silly to have to spell out the obvious things like "It should be a tactical FPS", so I thought I'd give my list as a list of things I'd most want to improve compared to the original Ghost Recon. Since I'm guessing that's the game that SG/GB will be most similar to. So this is my list, in approximate order, I've tried to keep it to just 10 items Random placement of enemy (with lots of possible spawn points) and random patrol routes. Scalable difficulty: being able to vary the number of enemies. For the AI to behave more realistically in CQB For the AI to be better at seeking cover and concealment when threatened. For the AI to react realistically to suppression fire and for them to try and suppress the players. For the AI spotting and alertness behaviour to be more realistic, including how they see through concealment. More realistic weapon modelling (move and fire) and better weapon and loadout customization. Flashbangs and smoke grenades (yeah OGR never had them, very odd!) More detailed maps, with more cover/conceal and longer view distance. Better graphics generally (characters, enemy, weapons, vehicles) As I say, I'd be really interested to know what are other people's most important features
  14. Aim convergence

    I agree completely on artificial balancing That said the real world has limits too. Weight and length matter in firearms, otherwise why have SMGs at all, why not just use M60s for CQB? An M60 has a great ammo capacity, good recoil and a nice high powered round. Because differences in weight and length matter in the real world these differences have to be modelled in the game as well. Sure you can spin a SAW around fairly quickly onto a new target, but it shouldn't be as quickly as a SMG. If you can spin and reaim a SAW as fast as an SMG, then why include SMGs in the game at all? I'm very confident that BFS will balance the weapons in a way that is accurate and not at all artificial. The question I'm interested in here is looking at what kind of model would allow them to do this best. Computer games are after all nothing more than models, and the only way to get realistic behaviour in the game is to come up with a model that behaves realistically
  15. Aim convergence

    I'm very hopefully also, I think BFS have really got a good focus. Still I'd be really interested to know whether this issue has been considered I'd also be interested what the community thinks