EPTOn March 07 2013 08:08 mutantmagnet wrote:
I don't know what's sillier, a scientific study, that probably cost thousands of dollars, to point out that Terrans need to win Mid Game or the get steam rolled; when after 2 years of community collaboration we had already figured this out with less money directly involved, or that a bunch of people who can carefully analyze build orders can't carefully read a very brief statistical report.
I don't know what's sillier, a scientific study, that probably cost thousands of dollars, to point out that Terrans need to win Mid Game or the get steam rolled; when after 2 years of community collaboration we had already figured this out with less money directly involved, or that a bunch of people who can carefully analyze build orders can't carefully read a very brief statistical report.
This article doesn't explain it very well, but the study does NOT look at what race is more likely to win a game of SC2.
The study says, assume a galaxy with stars and planets distributed like the Milky Way, on some given planet is one of the 3 races, these races fight for control of other planets and are successful with the probability of winning in a game of SC2, then what race will eventually dominate the galaxy?
The aim of the study was not to study SC2 win rates or winning strategies, it was to simulate "interstellar colonisation".
To investigate how the StarCraft civilisations would fare in a realistic
Galactic environment, we carry out Monte Carlo Realisation
(MCR) simulations of a region of the Milky Way. Each realisation
is an independent simulation, where the six civilisations compete
to colonise all stars in the region. As the simulation evolves, the
civilisations expand their occupied regions, come into contact, and
fight for the ownership of star systems. Running multiple MCRs
allows us to characterise the random error in this process (see e.g.
Vukotic & Cirkovic 2007, Forgan 2009). The MCRs were carried
out in a cubic domain, with Nstar = 106 stars placed randomly according
to a constant stellar number density of 0.14pc−3. This set
the box length to be 191 pc (or around 620 light years). A neighbour
sphere was constructed for each star of radius 5 pc. All stars
found within this radius are added to a neighbours list - for a constant
stellar density, this corresponds to a typical neighbour count of
73. However, while the overall density is uniform, there are small
overdensities and underdensities present in the domain. This will
become important for civilisations attempting rapid colonisation.
In each realisation, one civilisation of each type was randomly
seeded onto a star. Civilisations then select the nearest star in that
star’s neighbour list unoccupied by their own species, and travel
towards it at fixed velocity. When a civilisation colonises a new
star, the civilisation must wait for a fixed time interval before initiating
a new colonisation mission, to represent the building time
required. Only one colonisation mission is allowed at any time for
any one star. The simulation selects a time-step according to the
smallest time interval required for any one colonisation mission
to arrive at its destination. Colonisation missions that arrive at a
planet occupied by another species engage in combat. The attacking
civilisation has a probability of victory against the defender
specified by the StarCraft 2 user data. This probability is used to
decide the victor. If the attacker wins, they assume control of the
star system and begin building a new colonisation mission. If the attacker
loses, their colonisation mission is destroyed, and the attackersmust
launch a new colonisation mission from their original location.
This process of initial expansion, contact with opponents, and
eventual total occupancy of the model region is shown in Figure 1.
Paper: http://www.roe.ac.uk/~tat/StarCraftScience_files/StarCraftScience.pdf
Galactic environment, we carry out Monte Carlo Realisation
(MCR) simulations of a region of the Milky Way. Each realisation
is an independent simulation, where the six civilisations compete
to colonise all stars in the region. As the simulation evolves, the
civilisations expand their occupied regions, come into contact, and
fight for the ownership of star systems. Running multiple MCRs
allows us to characterise the random error in this process (see e.g.
Vukotic & Cirkovic 2007, Forgan 2009). The MCRs were carried
out in a cubic domain, with Nstar = 106 stars placed randomly according
to a constant stellar number density of 0.14pc−3. This set
the box length to be 191 pc (or around 620 light years). A neighbour
sphere was constructed for each star of radius 5 pc. All stars
found within this radius are added to a neighbours list - for a constant
stellar density, this corresponds to a typical neighbour count of
73. However, while the overall density is uniform, there are small
overdensities and underdensities present in the domain. This will
become important for civilisations attempting rapid colonisation.
In each realisation, one civilisation of each type was randomly
seeded onto a star. Civilisations then select the nearest star in that
star’s neighbour list unoccupied by their own species, and travel
towards it at fixed velocity. When a civilisation colonises a new
star, the civilisation must wait for a fixed time interval before initiating
a new colonisation mission, to represent the building time
required. Only one colonisation mission is allowed at any time for
any one star. The simulation selects a time-step according to the
smallest time interval required for any one colonisation mission
to arrive at its destination. Colonisation missions that arrive at a
planet occupied by another species engage in combat. The attacking
civilisation has a probability of victory against the defender
specified by the StarCraft 2 user data. This probability is used to
decide the victor. If the attacker wins, they assume control of the
star system and begin building a new colonisation mission. If the attacker
loses, their colonisation mission is destroyed, and the attackersmust
launch a new colonisation mission from their original location.
This process of initial expansion, contact with opponents, and
eventual total occupancy of the model region is shown in Figure 1.
Paper: http://www.roe.ac.uk/~tat/StarCraftScience_files/StarCraftScience.pdf
