There aren’t many things that matter for high MPG in a conventional (non-hybrid) car.
Modern engines are all pretty close efficiency-wise - a 302ci V8 replacing a 121ci inline 4 in the same car will not change the gas mileage*. Therefore, the chassis matters a lot more than the engine.
The figure that matters most for mileage in the city is curb weight. A lighter car will almost always get better mileage than a heavier one. In the city, almost all of what the engine does gets undone by the brakes - and what isn’t braked off is lost in rolling resistance and driveline friction, both of which depend primarily on weight. KE= 1/2 mv^2 so the more mass the car has the more energy you have to use to accelerate it to the same speed and the more energy you lose when you stop. Pumping losses in an automatic transmission screw you over in the city; I’ve never seen an automatic car achieve its EPA city mileage rating in actual city use whereas the mileage estimate isn’t too far off on manuals.
Out on the highway, weight becomes slightly less important; aerodynamics and gearing play starring roles here. The slower the engine has to turn on the highway, the better. “Longer” gearing reduces the engine’s pumping and frictional losses of the engine. This is why manual-transmission Corvettes get more than 35 MPG just cruising down the highway; that big motor’s barely turning above idle. How much your mileage depends on speed generally depends on how aerodynamically slick it is. There are SUVs that burn twice as much gas going 75 as they do going 55, but my old CRX gets the same gas mileage doing 85 as it does going 55 **. The EPA numbers are effectively worthless as the EPA highway test never goes over 48 MPH - if you put a Corvette into 6th at 48 the engine would be shaking like a leaf as the computer tried to prevent it from stalling.
As for how energetically a car accelerates at normal speeds, there are exactly three things that matter: power-to-weight ratio, gearing, and driver ability.
By far the biggest factor is the power-to-weight ratio. It would be nearly impossible to get a car that has 10 lbs/horsepower to feel slow in everyday driving; it would be nearly impossible to get a car that has 30 lbs/hp to feel fast. Most cars are (thankfully) closer to the former than the latter nowadays.
Torque really doesn’t matter - you just need to rev the engine higher to get the same performance out of an engine with the same power output and less torque. This typically means being in a lower or shorter gear. A 240-horsepower, 2800 lb car will achieve the same performance regardless of whether it has 150 lbsft of torque (Honda S2000) or 300 lbsft of torque ('93 Ford Mustang Cobra) - the Honda’s engine just has to turn nearly twice as fast as the Ford’s.
Most cars are geared pretty well for both acceleration and economy nowadays. With the six-speed manual gearbox fast becoming the standard, you can have some very short first and second gears to shove you out of the blocks hard while still having a sixth gear long enough to turn 2500 RPM at 70.
Most drivers are wasting their money on extra power that they never use. If you never put the throttle to the floor and don’t rev the engine to near the redline every time you go to merge onto a highway or pass somebody, then you’re not using all the power you paid for and would’ve been better off going with a less-powerful version of the same car - a four instead of a six, the 120 horsepower 120ci four instead of the 160 horsepower version of the same engine. Don’t worry about wearing the engine out - just about anything that isn’t a Ford should be able to make it to 200,000 miles without an overhaul.
So, to sum up, just get the lightest car you can find, with a six-speed manual if you can find it and a five-speed if you can’t, and then spec out as much power as you’ll use - 95% of the time it won’t matter economy-wise and during the other 5% economy’s the last thing on your mind.
*This is based on observed mileage of Ford Focuses with 302 engine swaps versus previous mileage by the same driver in the same car. This repeats itself for swaps that don’t involve such major surgery - a hot B-series engine in a Civic will typically get better gas mileage than the original D-series motor despite (usually) additional displacement and up to twice the power.
**As an engineer, it distresses me somewhat to write this. There are a lot of logical reasons for a car’s mileage to drop off at higher speeds: the aerodynamic drag force goes up with the square of the speed, the frictional drag force goes up linearly with speed, and the engine has to turn faster so its pumping losses increase.
However, I like empirical data - it’s really what matters at the end of the day. There is one huge reason an aerodynamically slick car gets virtually identical real-world highway mileage over a wide range of speeds. You don’t have as wide a speed variation over hills at high speeds - on an uphill grade where you have to floor it to keep the car going 55 in high gear or downshift, you might not even have to open the throttle any more to climb the same hill at 85. Throttle and speed variations are even worse for mileage than just a steady fast pace.
In addition, automakers program injection computers to 1) ace EPA mileage and emissions tests and 2) keep the engine from being damaged during full-throttle abuse by car magazines. The throttle is never floored during an EPA test. As a consequence, there is absolutely no official penalty for a car whose computer tells the injectors to just dump fuel into the engine under full throttle in order to avoid detonation. When journalists fill the tank with the rattiest 87 octane fuel they can find, go out to a SoCal track on a 100-degree day, and pound on the car for hours on end, the last thing the automakers want to hear is that the magazine blew the car up. As a consequence, the mixture is made a lot richer at full throttle than it needs to be - extra fuel decreases combustion temperatures and evaporates on the intake stroke, taking away some more heat when it does.