*Suggestion* New Tech Algorithm

[John DiFool the 2nd]

[Warning long sleep-inducing post!]

I've been concerned about the large effect lucks has on Research since SC1, as have many of you. The following is a suggestion of how the effect of luck can be minimized while still facilitating replayability.

As It Is Now

Each turn, you have a 5/4/3/2/1% chance per chit invested of having a tech pay off. Why is this bad? Because the chit can pay off immediately, or a longggg ways down the road.

This leads to gameplay imbalances and frustration on the part of the player(s).

My Proposal

The following exercise will, for simplicity, assume the following:

1. The tech in question is at 0 level (hence 5% per chit chance of advancement).

2. Intel and "catch up" play no part in the calculations.

3. There aren't any "hidden" algorithms already in place within the tech subroutines which alter how tech works "according to the manual."

The Non-Technical Description

Tech advances will proceed on a scale of 0-100, with 0 equating to the start of a new research cycle, and 100 equating to a new tech advancement.

Each turn, according to the number of chits invested (plus the other accompanying factors), the research level will increase on the 1-100 scale by some semi-random amount.

Example: 4 chits are invested in Advanced Subs, which at the moment is at the lowest (0) level, for a total expected (average) percentage gain per turn of 20% (4x5%). The research level will thus advance by 1-40 points on the 1-100 scale; lowest possible being 1 point, highest possible being double the current percentage of 20%, or 40.

So the cumulative progression might look like this (straight off my programmable calculator):

Turn 1: 6

Turn 2: 13

Turn 3: 24

Turn 4: 39

Turn 5: 78 [expected # of turns]

Turn 6: 117 - Advance!

As you can see I was unlucky at the beginning, but

got 2 39 point increases at the end.

Technical Description (feel free to skip)

The algorithm I used is as follows:

Current Level (0-100) + Random # from 1 to [ 2 x Chits x Base % ] = New Level

I also played around with one with a NEGATIVE increase (representing dead-end research alleys), which actually is the one I used in the table below:

CL (0-100) + Random # from [-1 x Chits] to [2.2 x Chits x Base %]

Hence in the example above (4 chits @ level 0) we have:

A random number from -4 to 44.

Results

For time constraints (this is on my lunch hour), I only simulated w/ chits of 1, 3 and 5, for 500 trials each (i.e. the time to get a new advance was 1 trial, whether it took 1 turn or 141 turns, as happened in one trial I did!).

Hubert's Way, followed by my way

H: 1 Chit: Mean 19.8 turns (expected 20.0 turns), Standard Deviation 17.9 turns

Me: 1 Chit: Mean 20.78 turns, SD 3.57

H: 3 Chits: Mean 6.63 (exp. 6.7 turns), SD 5.84

M: 3 Chits: Mean 7.27, SD 1.87

H: 5 Chits: Mean 4.12 (exp. 4.00 turns), SD 3.62

M: 5 Chits: Mean 4.68, SD 1.54

As you can see, while Hubert's was close to the expected # of turns for advancement, but his standard deviation was all over the map. For example, in approximately 5% of the trials the advancement took 3x as long or more than expected (i.e. with 5 chits 12+ turns instead of 4), which to me (IMHO) is intolerable. This included a wait of 141 turns for one of the trials w/ one chit. None of my trials ever approached that 3x level. For example, using just 100 trials each, we respectively got (w/ 5 chits):

Turns to Advance-Hubert (out of 100)

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

1 21

2 18

3 9

4 10

5 5

6 8

7 7

8 7

9 6

10 4

11 1

12+ 4 (including one of 19 turns!)

Turns to Advance-My Method (out of 100)

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

1 0

2 3

3 14

4 32

5 29

6 13

7 9

8+ 0

Hubert's method then has 2 problems: it can pay off immediately (which is nonsensical) or it can pay off far too down the road to make a difference for the amount invested. My way, while still incorporating random factors, is quite a bit more predictable WITHOUT being completely predictable. A majority of the results are clustered around the expected payoff time of 4 turns, with a few outliers. You will have a definite wait before the expected payoff, and will have to plan for that.

You could, if you like, make the advancement levels known to the players (not to the enemy of course unless Intel gives them a hint), or keep them secret. I'll shut up now because this is too long already and I now await feedback, from the two of you out there who understood my rantings. :cool:

[Lars]

Why is an immediate payoff nonsensical?

It's not all rocket science.

[John DiFool the 2nd]

Because you don't decide one week that, "Men, today we embark on a top secret program to create a better bomber!" and 2 weeks later they start rolling off the assembly lines.

[Lars]

No, but you can decide to hang a pre-existing 75mm gun off a pre-existing chassis.

Btw, for the bomber, it's more a matter of deciding to build four-engine planes instead of two-engine ones.

[TaoJah]

I never liked random numbers, so I like the proposal. And science like this is pretty predictable : you know that when you invest x scientists for y months, you'll get an airplane to go further.

[Lars]

I'd say it depends more on the field and the country doing the research. It's not a bad idea, it's just you can't apply it across the board.

[StefMan]

Brilliant. Really like that idea. Great work John DiFool.

Of course it is nice if you get lucky in a game and get tech advances while your opponent invests just as much and keeps lagging behind - but the same thing really sucks when you are the one who has 5 chits invested and keeps lurking around level 1 or 2.

Keeping standard deviation within a reasonable range is just fair. Still some chances to get lucky be no more frustrating research droughts.

[Lars]

I've found upgrading all your units is a really great way to get Production Research the next turn.

[LampCord]

Looks good to me. Nice grouping within the expected timeframe but still not deterministic.

[Moonslayer]

The idea sounds appealling at first but I can see things getting mighty predictable game after game. Personally I like random... it adds an edge of excitement and simulates the real situation much more closely imho.

Good work on the calculations... but I find the spread of possible completion times far too restrictive to my liking. I prefer the current method.

[Sombra]

Originally posted by TaoJah:

I never liked random numbers, so I like the proposal. And science like this is pretty predictable : you know that when you invest x scientists for y months, you'll get an airplane to go further.

You don´t work in R&D right?

I like the research system at least basically. I would like to see a better catch up, some kind of general defense tech and more use of rockets perhaps some rebalancing along the way.

[Lars]

John did leave out a major player.

2. Intel and "catch up" play no part in the calculations.

Run the numbers again, and consider the effect of going from Intel 0 and gradually (or quickly, depending on the breaks) progressing to Intel 5.

Think that'll shift the numbers more to his liking.

[SeaMonkey]

Looking at your numbers John, it seems that we will always see an advancement within 7 turns max by your system. Half the time, 50 percentile, we will see that advancement by the fourth turn.

This of course assumes a five chit investment in that category.

I'm concluding correctly?

What I particularly like is the minimum turn deviation window for your proposal, especially for 1 chit.

One thing is apparent. If we want an advancement with the current model, we better invest at least 3 chits, a one to twelve turn usual occurrence.

Now the final straw, are we positive that either of these models is better than the other at accurately portraying research advancements as they happen IRL?

[Timskorn]

I can definitely see the argument against the current research model. It doesn't make sense that upon researching a new tech it's available a week later, not for everything anyway. Or how about the fact that my armor 1000 miles east of Germany can then also upgrade all of their tanks to the latest model in a single week?

My point is, the current system COULD be more realistic, but I like it the way it is. It adds unpredictable variables into a game that already runs pretty predictably. Axis attacks Poland, then Denmark, then France, then Norway, then Yugo, then Russia...for the majority of instances this is how it goes. I'm going to play this game enough times to want to see some pretty different variations. I mean one game I might get heavy tank level 3 by 1941, in another not until 1943. Will it make or break my game? Possibly, but to me that's a challenge to overcome.

[John DiFool the 2nd]

Yes SM, but that will drop as the tech levels go up

(the 5/4/3/2/1 system). At lower percentages the

spread gets larger (as my data indicates, for both

systems), so in the long term that isn't a problem.

I do need to sim a entire research program, from level

0 to level 5...

This is just a starting point mind you, something to

get the discussion going; I could see the chances

"spread out" more if necessary.

Short reply to the point Timskorn raised: should

variation in tech advances be driven by the random

seed, or our choices (or a mixture of both)? I

doubt many SCers would be as stupid as Hitler, who

only geared up his heavy tank research after the

Wehrmacht got surprised by T-34s and KV-1s in

1941. I still think there's enough choice in the

system now (definitely more than in SC1) that

"cookie cutter" tech investments using something

like my system won't necessarily lead to predictable

outcomes down the road.

[Ottosmops]

What about an initial research delay for a few turns?

Regardless how many chits you have invested, for the first 3 ( or 5 or 7? ) turns you can't advance to a new level.

This could work with the current system and also with John DiFool's system.

[Desert Dave]

A... Gathering of Geeks

Some would call it, but,

Not I.

No, never, not I.

Maybe back in the ram-rod days

Of riding the BSA or Indian

All over the high mesa,

And through rutted flash-flood gullies,

But, no, nix, not now.

NO WAY!

Due to the following

Intractable, in factual didactum:

In the next year or 2,

Chinese institutes of "higher learning"

Will graduate - get this! Math Cats!

650,000 Engineers!

India will emit forth

From geek institutes... 350,000 Engineers!

Make that... total of: ONE MILLION

For them Old Kingdom kind of Countries.

US of A?

Squabbling like 2-year olds

Over useless and stupid phenomenon

Like Who! Can horde the most Lucre?

Which they can't take with 'em

When they kick the bucket?

A grand total of... 30,000 Engineers.

Well, hell's bell's, I wonder who?

Will be a'building

Eiffel-like edifices & cheap labor

Sweat shops - this next... Century?

My own son is an Engineer.

Hope he don't have to move in with me

Anytime soon.

I already raised him... the once.

____________________________________

JdF2, unsure of what might come next:

I'll shut up now because this is too long already and I now await feedback, from the two of you out there who understood my rantings.

I would say... Moonslayer & Timskorn!

Since the system as is - is,

IMHO, the very best one.

You CAN "catch-up" fairly quickly,

Given the "Law of Averages."

It's why we play the games, eh?

To - eventually, realize

The law of averages.

Who would want the SAME ravening & unraveling

Of Europe, every... single... game?

Or,

You could assist yerself, not insignificantly,

By investing in... "Intel."

Like the London Werewolf's

Slicked back hair,

I think what we got now

Is already... per-fect! :cool:

[Lars]

Originally posted by John DiFool the 2nd:

I

doubt many SCers would be as stupid as Hitler, who

only geared up his heavy tank research after the

Wehrmacht got surprised by T-34s and KV-1s in

1941.

That only seems stupid in hindsight.

If you've just crushed all of western Europe with pzIII and pzIV, and didn't know about T-34's, why bother working too hard on pzV and pzIV?

[Timskorn]

John, I like your system, it's very well thought out and would work fine I think as it still offers a good randomness factor, but a little more predictable. As a player I would always 'feel' like I'm just about to make a breakthrough in technology with each successive turn...if I knew the formula (or had played a lot).

Right now it feels very random because it is, but again I make a case for Hubert's system simply from a replayability sense. With your system we can reasonably expect a breakthrough within a certain number of turns, but with Hubert's we simply don't know. The difference can be a lot in terms of how we play the game.

As in most game design, this is all "Theorycraft" right now. Your formula would need to be played and tested by players to see how it affects gameplay, and whether it ultimately makes it better or worse than Hubert's system.

[Blashy]

I support the current system. Because some games you ARE unlucky while others you are VERY lucky.

It does not occur often, but enough to show that R&D breakthroughs or lack of it can be a game breaker, and it should be.

This "chance" on tech also simulates spying, if you have level 4 tanks and your opponent only has level 2, well maybe the codebreakers are having a rough time or your spies have been captures.

I like what Hubert has done, it is very exciting when in one game your opponent just won't get that level 3 IW and you've had it for 2 years, hehe.

They do not occur often, but just enough to no longer make SC2 like SC, same old same old way of playing the game.

[Desert Dave]

Replayability is paramount

Wouldn't you think?

Given the number of games you'll play

Over the course of a decade?

Guess it really comes down to:

Do you want to roll out

A sweaty hand-full of dice?

Or,

Rely on a "predetermined set"

Of very dry, continually

PREDICTABLE numbers?

Do you want to tangle with tremendous

Amounts of strange, odd, random

And astonishing! Surprises?

Or,

KNOW all the roads and signs, in advance?

Either way can be good,

Depending on what your compelling

Inclination is.

Seeing the new and unusual flower bloom?

Or,

Sitting in well-tended garden and noticing

The exact same pattern,

Day after day after day?

Like I say, to each their own.

__________________________________________

Besides,

Without Lady Luck the vamp traipsing through

The very orderly living-room,

Then, nobody could claim

They REALLY and truly lost the game! :eek:

Because

She has it in for them!

LOL!

[ April 21, 2006, 07:44 AM: Message edited by: Desert Dave ]

[Desert Dave]

You know,

There IS one way that would

Make your game somewhat less antic.

Simply use the Editor

And place a certain number of chits

In each research category.

No further investments allowed, AT ALL.

[... nor in Intel, if you really

want to narrow the range]

That way,

A particular Nation

Could ONLY reach a "fairly historical"

Level in each area.

You'd still have a little randomness,

But,

You would at least eliminate

Those advances which were not very likely

For some countries to achieve.

[ April 21, 2006, 08:21 AM: Message edited by: Desert Dave ]

[John DiFool the 2nd]

Well, I'll say one more thing and then I'll shut up:

if the "replayability" of SC2 depends mostly on tech

rolls (which is what everyone seems to be arguing),

then there should be improvement in terms of

replayability in other areas of the game system.

[Could be more blunt and unkind on this point]

For example why SHOULD there be Iraq and Yugo coups

every single game, at around the same time every

single game, while tech progresses on its own

merry unpredictable path? Couldn't you put some

percentages on those not happening at all,

along with some (small) chances of coups in

Turkey or revolutions in any of the Vichy holdings?

For that matter why not start the game in 1936,

with nobody at war (yes I know this can't be modded

right now because you need at least 2 major nation

combatants), and let events and choices determine

the shape of the future. Anyone who remembers

"Days of Decision" from World in Flames, or even

"Dice of Decision" from Totaler Krieg! know what

I mean. :cool:

[Desert Dave]

Well, I'll say one more thing and then I'll shut up: if the "replayability" of SC2 depends mostly on tech rolls (which is what everyone seems to be arguing), then there should be improvement in terms of replayability in other areas of the game system.

[Could be more blunt and unkind on this point]

JdF2,

Valid point.

As it was, we were ONLY discussing tech variability.

My opinion is MERELY: one of the wargame playing guys who's been around here for 4 years as of tomorrow.

However, there ARE other ways, IE, via "scripting" of events and AI actions/reactions, that would add some more unpredictability.

No need to shut up however; why ever?

Your points about this issue or ANYTHING else are equally as good as mine, or anyone else's.

I was "joshing around" with my comment about "geeks."

My apologies if it sounded as though I were making fun of you, or your proposal.

After all, my Son is an Engineer and I would NEVER consider him a geek.

You can be blunt as need be... no doubt Hubert is interested in ANY KIND of improvements that are commonly, consistently requested.

He has already proved that to be true in the past; QUITE unlikely he would change now.

Be forthcoming with those areas that you feel might be "gamey" or "exploitive," as well! :cool:

[Timskorn]

You're right John, which is what I was talking about when I said the 1939 game generally always follows a certain predictable path. However, thankfully we DO have the editor which can change a ton of things. The Iraq coup can be set to trigger only 75% of the time, for example, so it's a little more random. Same with the Yugo coup. I'm actually all for making things like this a little more random.