diff --git a/egui/src/input.rs b/egui/src/input.rs
index 7311d652..630edaff 100644
--- a/egui/src/input.rs
+++ b/egui/src/input.rs
@@ -244,6 +244,18 @@ impl InputState {
             )
         })
     }
+
+    /// Size of a physical pixel in logical gui coordinates (points).
+    pub fn physical_pixel_size(&self) -> f32 {
+        1.0 / self.pixels_per_point
+    }
+
+    /// How imprecise do we expect the mouse/touch input to be?
+    /// Returns imprecision in points.
+    pub fn aim_radius(&self) -> f32 {
+        // TODO: multiply by ~3 for touch inputs because fingers are fat
+        self.physical_pixel_size()
+    }
 }
 
 impl MouseInput {
diff --git a/egui/src/math.rs b/egui/src/math.rs
index 2b406944..4a48d681 100644
--- a/egui/src/math.rs
+++ b/egui/src/math.rs
@@ -7,6 +7,7 @@ use std::ops::{Add, Mul, RangeInclusive};
 mod movement_tracker;
 mod pos2;
 mod rect;
+pub mod smart_aim;
 mod vec2;
 
 pub use {movement_tracker::*, pos2::*, rect::*, vec2::*};
diff --git a/egui/src/math/smart_aim.rs b/egui/src/math/smart_aim.rs
new file mode 100644
index 00000000..4f857cbb
--- /dev/null
+++ b/egui/src/math/smart_aim.rs
@@ -0,0 +1,160 @@
+#![allow(clippy::float_cmp)] // I know what I'm doing
+
+const NUM_DECIMALS: usize = 15;
+
+pub fn best_in_range_f32(min: f32, max: f32) -> f32 {
+    best_in_range_f64(min as f64, max as f64) as f32
+}
+
+/// Find the "simplest" number in a closed range [min, max], i.e. the one with the fewest decimal digits.
+///
+/// So in the range `[0.83, 1.354]` you will get `1.0`, and for `[0.37, 0.48]` you will get `0.4`.
+/// This is used when dragging sliders etc to get the values that users are most likely to desire.
+/// This assumes a decimal centric user.
+pub fn best_in_range_f64(min: f64, max: f64) -> f64 {
+    // Avoid NaN if we can:
+    if min.is_nan() {
+        return max;
+    }
+    if max.is_nan() {
+        return min;
+    }
+
+    if max < min {
+        return best_in_range_f64(max, min);
+    }
+    if min == max {
+        return min;
+    }
+    if min <= 0.0 && 0.0 <= max {
+        return 0.0; // always prefer zero
+    }
+    if min < 0.0 {
+        return -best_in_range_f64(-max, -min);
+    }
+
+    // Prefer finite numbers:
+    if !max.is_finite() {
+        return min;
+    }
+    debug_assert!(min.is_finite() && max.is_finite());
+
+    let min_exponent = min.log10();
+    let max_exponent = max.log10();
+
+    if min_exponent.floor() != max_exponent.floor() {
+        // pick the geometric center of the two:
+        let exponent = (min_exponent + max_exponent) / 2.0;
+        return 10.0_f64.powi(exponent.round() as i32);
+    }
+
+    if is_integer(min_exponent) {
+        return 10.0_f64.powf(min_exponent);
+    }
+    if is_integer(max_exponent) {
+        return 10.0_f64.powf(max_exponent);
+    }
+
+    let exp_factor = 10.0_f64.powi(max_exponent.floor() as i32);
+
+    let min_str = to_decimal_string(min / exp_factor);
+    let max_str = to_decimal_string(max / exp_factor);
+
+    // eprintln!("min_str: {:?}", min_str);
+    // eprintln!("max_str: {:?}", max_str);
+
+    let mut ret_str = [0; NUM_DECIMALS];
+
+    // Select the common prefix:
+    let mut i = 0;
+    while i < NUM_DECIMALS && max_str[i] == min_str[i] {
+        ret_str[i] = max_str[i];
+        i += 1;
+    }
+
+    if i < NUM_DECIMALS {
+        // Pick the deciding digit.
+        // Note that "to_decimal_string" rounds down, so we that's why we add 1 here
+        ret_str[i] = simplest_digit_closed_range(min_str[i] + 1, max_str[i]);
+    }
+
+    from_decimal_string(&ret_str) * exp_factor
+}
+
+fn is_integer(f: f64) -> bool {
+    f.round() == f
+}
+
+fn to_decimal_string(v: f64) -> [i32; NUM_DECIMALS] {
+    debug_assert!(v < 10.0, "{:?}", v);
+    let mut digits = [0; NUM_DECIMALS];
+    let mut v = v.abs();
+    for r in digits.iter_mut() {
+        let digit = v.floor();
+        *r = digit as i32;
+        v -= digit;
+        v *= 10.0;
+    }
+    digits
+}
+
+fn from_decimal_string(s: &[i32]) -> f64 {
+    let mut ret: f64 = 0.0;
+    for (i, &digit) in s.iter().enumerate() {
+        ret += (digit as f64) * 10.0_f64.powi(-(i as i32));
+    }
+    ret
+}
+
+/// Find the simplest integer in the range [min, max]
+fn simplest_digit_closed_range(min: i32, max: i32) -> i32 {
+    debug_assert!(1 <= min && min <= max && max <= 9);
+    if min <= 5 && 5 <= max {
+        5
+    } else {
+        (min + max) / 2
+    }
+}
+
+#[test]
+fn test_aim() {
+    assert_eq!(best_in_range_f64(-0.2, 0.0), 0.0, "Prefer zero");
+    assert_eq!(best_in_range_f64(-10_004.23, 3.14), 0.0, "Prefer zero");
+    assert_eq!(best_in_range_f64(-0.2, 100.0), 0.0, "Prefer zero");
+    assert_eq!(best_in_range_f64(0.2, 0.0), 0.0, "Prefer zero");
+    assert_eq!(best_in_range_f64(7.8, 17.8), 10.0);
+    assert_eq!(best_in_range_f64(99.0, 300.0), 100.0);
+    assert_eq!(best_in_range_f64(-99.0, -300.0), -100.0);
+    assert_eq!(best_in_range_f64(0.4, 0.9), 0.5, "Prefer ending on 5");
+    assert_eq!(best_in_range_f64(14.1, 19.99), 15.0, "Prefer ending on 5");
+    assert_eq!(best_in_range_f64(12.3, 65.9), 50.0, "Prefer leading 5");
+    assert_eq!(best_in_range_f64(493.0, 879.0), 500.0, "Prefer leading 5");
+    assert_eq!(best_in_range_f64(0.37, 0.48), 0.40);
+    // assert_eq!(best_in_range_f64(123.71, 123.76), 123.75); // TODO: we get 123.74999999999999 here
+    assert_eq!(best_in_range_f32(123.71, 123.76), 123.75); // TODO: we get 123.74999999999999 here
+    assert_eq!(best_in_range_f64(7.5, 16.3), 10.0);
+    assert_eq!(best_in_range_f64(7.5, 76.3), 10.0);
+    assert_eq!(best_in_range_f64(7.5, 763.3), 100.0);
+    assert_eq!(best_in_range_f64(7.5, 1_345.0), 100.0);
+    assert_eq!(best_in_range_f64(7.5, 123_456.0), 1000.0, "Geometric mean");
+    assert_eq!(best_in_range_f64(9.9999, 99.999), 10.0);
+    assert_eq!(best_in_range_f64(10.000, 99.999), 10.0);
+    assert_eq!(best_in_range_f64(10.001, 99.999), 50.0);
+    assert_eq!(best_in_range_f64(10.001, 100.000), 100.0);
+    assert_eq!(best_in_range_f64(99.999, 100.000), 100.0);
+    assert_eq!(best_in_range_f64(10.001, 100.001), 100.0);
+
+    use std::f64::{INFINITY, NAN, NEG_INFINITY};
+    assert!(best_in_range_f64(NAN, NAN).is_nan());
+    assert_eq!(best_in_range_f64(NAN, 1.2), 1.2);
+    assert_eq!(best_in_range_f64(NAN, INFINITY), INFINITY);
+    assert_eq!(best_in_range_f64(1.2, NAN), 1.2);
+    assert_eq!(best_in_range_f64(1.2, INFINITY), 1.2);
+    assert_eq!(best_in_range_f64(INFINITY, 1.2), 1.2);
+    assert_eq!(best_in_range_f64(NEG_INFINITY, 1.2), 0.0);
+    assert_eq!(best_in_range_f64(NEG_INFINITY, -2.7), -2.7);
+    assert_eq!(best_in_range_f64(INFINITY, INFINITY), INFINITY);
+    assert_eq!(best_in_range_f64(NEG_INFINITY, NEG_INFINITY), NEG_INFINITY);
+    assert_eq!(best_in_range_f64(NEG_INFINITY, INFINITY), 0.0);
+    assert_eq!(best_in_range_f64(INFINITY, NEG_INFINITY), 0.0);
+}
diff --git a/egui/src/widgets.rs b/egui/src/widgets.rs
index cc9b3c16..6b9cec11 100644
--- a/egui/src/widgets.rs
+++ b/egui/src/widgets.rs
@@ -581,6 +581,8 @@ impl<'a> Widget for DragValue<'a> {
             if delta_value != 0.0 {
                 *value += delta_value;
                 *value = round_to_precision(*value, precision);
+                // TODO: To make use or `smart_aim` for `DragValue` we need to store some state somewhere,
+                // otherwise we will just keep rounding to the same value while moving the mouse.
             }
         }
         interact.into()
diff --git a/egui/src/widgets/slider.rs b/egui/src/widgets/slider.rs
index 94b41b5b..48eacd12 100644
--- a/egui/src/widgets/slider.rs
+++ b/egui/src/widgets/slider.rs
@@ -82,6 +82,9 @@ impl<'a> Slider<'a> {
         self
     }
 
+    /// Precision (number of decimals) used when displaying the value.
+    /// Values will also be rounded to this precision.
+    /// Regardless of precision the slider will use "smart aim" to help the user select nice, round values.
     pub fn precision(mut self, precision: usize) -> Self {
         self.precision = precision;
         self
@@ -95,6 +98,11 @@ impl<'a> Slider<'a> {
         value = round_to_precision(value, self.precision);
         (self.get_set_value)(Some(value));
     }
+
+    /// For instance, `point` is the mouse position and `point_range` is the physical location of the slider on the screen.
+    fn value_from_point(&self, point: f32, point_range: RangeInclusive<f32>) -> f32 {
+        remap_clamp(point, point_range, self.range.clone())
+    }
 }
 
 impl<'a> Widget for Slider<'a> {
@@ -154,7 +162,12 @@ impl<'a> Widget for Slider<'a> {
 
             if let Some(mouse_pos) = ui.input().mouse.pos {
                 if interact.active {
-                    self.set_value_f32(remap_clamp(mouse_pos.x, left..=right, range.clone()));
+                    let aim_radius = ui.input().aim_radius();
+                    let new_value = crate::math::smart_aim::best_in_range_f32(
+                        self.value_from_point(mouse_pos.x - aim_radius, left..=right),
+                        self.value_from_point(mouse_pos.x + aim_radius, left..=right),
+                    );
+                    self.set_value_f32(new_value);
                 }
             }