diff --git a/src/yaz0.rs b/src/yaz0.rs
index 1e6d400..6123576 100644
--- a/src/yaz0.rs
+++ b/src/yaz0.rs
@@ -1,5 +1,7 @@
 // Based on https://gist.github.com/Mr-Wiseguy/6cca110d74b32b5bb19b76cfa2d7ab4f
 
+use std::cmp;
+
 use crate::utils;
 
 pub fn decompress_yaz0(bytes: &[u8]) -> Box<[u8]> {
@@ -60,6 +62,265 @@ pub fn decompress_yaz0(bytes: &[u8]) -> Box<[u8]> {
     ret.into_boxed_slice()
 }
 
-pub fn compress_yaz0(_bytes: &[u8]) -> Box<[u8]> {
-    panic!("Not implemented")
+fn divide_round_up(a: usize, b: usize) -> usize {
+    (a + b - 1) / b
+}
+
+pub fn compress_yaz0(bytes: &[u8]) -> Box<[u8]> {
+    let input_size = bytes.len();
+    // Worst-case size for output is zero compression on the input, meaning the input size plus the number of layout bytes plus the Yaz0 header.
+    // There would be one layout byte for every 8 input bytes, so the worst-case size is:
+    //   input_size + ROUND_UP_DIVIDE(input_size, 8) + 0x10
+    let mut output: Vec<u8> =
+        Vec::with_capacity(input_size + divide_round_up(input_size, 8) + 0x10);
+
+    output.extend(b"Yaz0");
+    output.extend((input_size as u32).to_be_bytes());
+    // padding
+    output.extend(&[0, 0, 0, 0, 0, 0, 0, 0]);
+
+    output.push(0);
+    let mut index_cur_layout_byte: usize = 0x10;
+    let mut index_out_ptr: usize = index_cur_layout_byte + 1;
+    let mut input_pos: usize = 0;
+    let mut cur_layout_bit: u8 = 0x80;
+
+    while input_pos < input_size {
+        let mut group_pos: i32 = 0;
+        let mut group_size: u32 = 0;
+
+        search(
+            input_pos,
+            input_size,
+            &mut group_pos,
+            &mut group_size,
+            bytes,
+        );
+
+        // If the group isn't larger than 2 bytes, copying the input without compression is smaller
+        if group_size <= 2 {
+            // Set the current layout bit to indicate that this is an uncompressed byte
+            output[index_cur_layout_byte] |= cur_layout_bit;
+            output.push(bytes[input_pos]);
+            input_pos += 1;
+            index_out_ptr += 1;
+        } else {
+            let mut new_size: u32 = 0;
+            let mut new_position: i32 = 0;
+
+            // Search for a new group after one position after the current one
+            search(
+                input_pos + 1,
+                input_size,
+                &mut new_position,
+                &mut new_size,
+                bytes,
+            );
+
+            // If the new group is better than the current group by at least 2 bytes, use it one instead
+            if new_size >= group_size + 2 {
+                // Mark the current layout bit to skip compressing this byte, as the next input position yielded better compression
+                output[index_cur_layout_byte] |= cur_layout_bit;
+                // Copy the input byte to the output
+                output.push(bytes[input_pos]);
+                input_pos += 1;
+                index_out_ptr += 1;
+
+                // Advance to the next layout bit
+                cur_layout_bit >>= 1;
+
+                if cur_layout_bit == 0 {
+                    cur_layout_bit = 0x80;
+                    index_cur_layout_byte = index_out_ptr;
+                    index_out_ptr += 1;
+                    output[index_cur_layout_byte] = 0;
+                }
+
+                group_size = new_size;
+                group_pos = new_position;
+            }
+
+            // Calculate the offset for the current group
+            let group_offset: u32 = (input_pos as i32 - group_pos - 1) as u32;
+
+            // Determine which encoding to use for the current group
+            if group_size >= 0x12 {
+                // Three bytes, 0RRRNN
+                output.push((group_offset >> 8) as u8);
+                index_out_ptr += 1;
+                output.push((group_offset & 0xFF) as u8);
+                index_out_ptr += 1;
+                output.push((group_size - 0x12) as u8);
+                index_out_ptr += 1;
+            } else {
+                // Two bytes, NRRR
+                output.push((group_offset >> 8) as u8 | ((group_size - 2) << 4) as u8);
+                index_out_ptr += 1;
+                output.push((group_offset & 0xFF) as u8);
+                index_out_ptr += 1;
+            }
+
+            // Move forward in the input by the size of the group
+            input_pos += group_size as usize;
+        }
+
+        // Advance to the next layout bit
+        cur_layout_bit >>= 1;
+
+        if cur_layout_bit == 0 {
+            cur_layout_bit = 0x80;
+            index_cur_layout_byte = index_out_ptr;
+            output.push(0);
+            index_out_ptr += 1;
+        }
+    }
+
+    output.shrink_to_fit();
+    output.into_boxed_slice()
+}
+
+fn search(
+    input_pos: usize,
+    input_size: usize,
+    pos_out: &mut i32,
+    size_out: &mut u32,
+    data_in: &[u8],
+) {
+    let mut cur_size: usize = 3;
+    let mut found_pos: isize = 0;
+    let mut search_pos: usize = cmp::max(input_pos as isize - 0x1000, 0) as usize;
+    let search_size = cmp::min(input_size - input_pos, 0x111);
+
+    if search_size >= 3 {
+        while search_pos < input_pos {
+            let found_offset = mischarsearch(
+                &data_in[input_pos..],
+                cur_size,
+                &data_in[search_pos..],
+                cur_size + input_pos - search_pos,
+            );
+
+            if found_offset >= input_pos - search_pos {
+                break;
+            }
+
+            while cur_size < search_size {
+                if data_in[cur_size + search_pos + found_offset] != data_in[cur_size + input_pos] {
+                    break;
+                }
+                cur_size += 1;
+            }
+
+            if search_size == cur_size {
+                *pos_out = (found_offset + search_pos) as i32;
+                *size_out = cur_size as u32;
+                return;
+            }
+
+            found_pos = (search_pos + found_offset) as isize;
+            search_pos = (found_pos + 1) as usize;
+            cur_size += 1;
+        }
+
+        *pos_out = found_pos as i32;
+        if cur_size > 3 {
+            cur_size -= 1;
+            *size_out = cur_size as u32;
+            return;
+        }
+    } else {
+        *pos_out = 0;
+    }
+    *size_out = 0;
+}
+
+fn mischarsearch(pattern: &[u8], pattern_len: usize, data: &[u8], data_len: usize) -> usize {
+    let mut skip_table = [0u16; 256];
+    let mut i: isize;
+    //let mut k: usize;
+
+    let mut v6: isize;
+    let mut j: isize;
+
+    if pattern_len <= data_len {
+        initskip(pattern, pattern_len as i32, &mut skip_table);
+
+        i = pattern_len as isize - 1;
+        loop {
+            if pattern[pattern_len - 1] == data[i as usize] {
+                i -= 1;
+                j = pattern_len as isize - 2;
+                if j < 0 {
+                    return (i + 1) as usize;
+                }
+
+                while pattern[j as usize] == data[i as usize] {
+                    i -= 1;
+                    j -= 1;
+                    if j < 0 {
+                        return (i + 1) as usize;
+                    }
+                }
+
+                v6 = pattern_len as isize - j;
+
+                if skip_table[data[i as usize] as usize] as isize > v6 {
+                    v6 = skip_table[data[i as usize] as usize] as isize;
+                }
+            } else {
+                v6 = skip_table[data[i as usize] as usize] as isize;
+            }
+            i += v6;
+        }
+    }
+    data_len
+}
+
+fn initskip(pattern: &[u8], len: i32, skip: &mut [u16; 256]) {
+    skip.fill(len as u16);
+
+    for i in 0..len {
+        skip[pattern[i as usize] as usize] = (len - i - 1) as u16;
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    #[test]
+    fn test_matching_decompression() {
+        let compressed_file = include_bytes!("../test_data/Yaz0/1.Yaz0");
+        let decompressed_file = include_bytes!("../test_data/Yaz0/1.bin");
+
+        let decompressed: Box<[u8]> = super::decompress_yaz0(compressed_file);
+        assert_eq!(decompressed_file, decompressed.as_ref());
+    }
+
+    #[test]
+    fn test_matching_compression() {
+        let compressed_file = include_bytes!("../test_data/Yaz0/1.Yaz0");
+        let decompressed_file = include_bytes!("../test_data/Yaz0/1.bin");
+
+        let compressed = super::compress_yaz0(decompressed_file.as_slice());
+        assert_eq!(compressed_file, compressed.as_ref());
+    }
+
+    #[test]
+    fn test_cycle_decompressed() {
+        let decompressed_file = include_bytes!("../test_data/Yaz0/1.bin");
+
+        assert_eq!(
+            decompressed_file,
+            super::decompress_yaz0(&super::compress_yaz0(decompressed_file.as_ref())).as_ref()
+        );
+    }
+
+    #[test]
+    fn test_cycle_compressed() {
+        let compressed_file = include_bytes!("../test_data/Yaz0/1.Yaz0");
+
+        assert_eq!(
+            compressed_file,
+            super::compress_yaz0(&super::decompress_yaz0(compressed_file.as_ref())).as_ref()
+        );
+    }
 }
diff --git a/test_data/.gitignore b/test_data/.gitignore
index ed45cff..bfd1ca9 100644
--- a/test_data/.gitignore
+++ b/test_data/.gitignore
@@ -1,2 +1,3 @@
-!*.Yay0
 !*.bin
+!*.Yay0
+!*.Yaz0
diff --git a/test_data/Yaz0/1.Yaz0 b/test_data/Yaz0/1.Yaz0
new file mode 100644
index 0000000..1270e8b
Binary files /dev/null and b/test_data/Yaz0/1.Yaz0 differ
diff --git a/test_data/Yaz0/1.bin b/test_data/Yaz0/1.bin
new file mode 100755
index 0000000..b9559a5
--- /dev/null
+++ b/test_data/Yaz0/1.bin
@@ -0,0 +1 @@
+ASCII文字と互換性を持たせるために、ASCIIと同じ部分は1バイト=asdfasdfasdfasdfasdfasdfasdfasdfasdf
\ No newline at end of file